Compare commits
13 Commits
c073d945d1
...
master
| Author | SHA1 | Date | |
|---|---|---|---|
| 8cba417317 | |||
| 369454f877 | |||
| 6a65ed51a5 | |||
| 572e1d5abd | |||
| 504f0d7935 | |||
| d424d56143 | |||
| f34ae9b8c4 | |||
| 4568a119a2 | |||
| 4769d4b57f | |||
| 616652f67a | |||
| 946ae5f977 | |||
| b1517f1c35 | |||
| 5592f7a37d |
@@ -1,6 +1,6 @@
|
||||
MIT License
|
||||
|
||||
Copyright (c) 2020 Dehaeze Thomas
|
||||
Copyright (c) 2021 Dehaeze Thomas
|
||||
|
||||
Permission is hereby granted, free of charge, to any person obtaining a copy
|
||||
of this software and associated documentation files (the "Software"), to deal
|
||||
|
||||
85
index.html
85
index.html
@@ -3,9 +3,9 @@
|
||||
"http://www.w3.org/TR/xhtml1/DTD/xhtml1-strict.dtd">
|
||||
<html xmlns="http://www.w3.org/1999/xhtml" lang="en" xml:lang="en">
|
||||
<head>
|
||||
<!-- 2021-09-01 mer. 11:36 -->
|
||||
<!-- 2021-09-02 jeu. 10:02 -->
|
||||
<meta http-equiv="Content-Type" content="text/html;charset=utf-8" />
|
||||
<title>Complementary Filters Shaping Using $\mathcal{H}_\infty$ Synthesis</title>
|
||||
<title>A new method of designing complementary filters for sensor fusion using the $\mathcal{H}_\infty$ synthesis</title>
|
||||
<meta name="author" content="Thomas Dehaeze" />
|
||||
<meta name="generator" content="Org Mode" />
|
||||
<link rel="stylesheet" type="text/css" href="https://research.tdehaeze.xyz/css/style.css"/>
|
||||
@@ -34,7 +34,7 @@
|
||||
|
|
||||
<a accesskey="H" href="../index.html"> HOME </a>
|
||||
</div><div id="content" class="content">
|
||||
<h1 class="title">Complementary Filters Shaping Using \(\mathcal{H}_\infty\) Synthesis
|
||||
<h1 class="title">A new method of designing complementary filters for sensor fusion using the \(\mathcal{H}_\infty\) synthesis
|
||||
<br />
|
||||
<span class="subtitle">Dehaeze Thomas, Verma Mohit and Collette Christophe</span>
|
||||
</h1>
|
||||
@@ -44,33 +44,88 @@
|
||||
</p>
|
||||
|
||||
<p>
|
||||
For many applications, large bandwidth and dynamic ranges are requiring to use several sensors, whose signals are combined using complementary filters.
|
||||
This paper presents a method for designing these complementary filters using \(\mathcal{H}_\infty\) synthesis that allows to shape the filter norms.
|
||||
This method is shown to be easily applicable for the synthesis of complex complementary filters.
|
||||
In order to obtain a better estimate of a quantity being measured, several sensors having different characteristics can be merged with a technique called “sensor fusion”.
|
||||
The obtained “super sensor” can combine the benefits of the individual sensors provided that the complementary filters used in the fusion are well designed.
|
||||
Indeed, properties of the super sensor are linked to the magnitude of the complementary filters.
|
||||
Properly shaping the magnitude of complementary filters is a difficult and time-consuming task.
|
||||
In this study, we address this issue and propose a new method for designing complementary filters.
|
||||
This method uses weighting functions to specify the wanted shape of the complementary filter that are then easily obtained using the standard \(\mathcal{H}_\infty\) synthesis.
|
||||
The proper choice of the weighting functions is discussed, and the effectiveness and simplicity of the design method is highlighted using several examples.
|
||||
Such synthesis method is further extended for the shaping of more than two complementary filters.
|
||||
</p>
|
||||
</blockquote>
|
||||
|
||||
<div id="outline-container-orgafee5c3" class="outline-2">
|
||||
<h2 id="orgafee5c3">Paper (<a href="journal/dehaeze21_desig_compl_filte.pdf">link</a>)</h2>
|
||||
<div class="outline-text-2" id="text-orgafee5c3">
|
||||
<div id="outline-container-org8d13d5d" class="outline-2">
|
||||
<h2 id="org8d13d5d">Journal Article (<a href="journal/dehaeze21_desig_compl_filte.pdf">link</a>)</h2>
|
||||
<div class="outline-text-2" id="text-org8d13d5d">
|
||||
<p>
|
||||
The paper has been created <a href="https://orgmode.org/">Org Mode</a> (generating <a href="https://www.latex-project.org/">LaTeX</a> code) under <a href="https://www.gnu.org/software/emacs/">Emacs</a>.
|
||||
</p>
|
||||
|
||||
<p>
|
||||
To cite this paper use the following BibTeX code.
|
||||
</p>
|
||||
<div class="org-src-container">
|
||||
<pre class="src src-bibtex"><span class="org-function-name">@article</span>{<span class="org-constant">dehaeze21_new_method_desig_compl_filter</span>,
|
||||
<span class="org-variable-name">author</span> = {Dehaeze, Thomas and Vermat, Mohit and Collette, Christophe},
|
||||
<span class="org-variable-name">title</span> = {A New Method of Designing Complementary Filters for Sensor
|
||||
Fusion Using the <span class="org-string">$\mathcal{H}_\infty$</span> Synthesis},
|
||||
<span class="org-variable-name">journal</span> = {Mechanical Systems and Signal Processing},
|
||||
<span class="org-variable-name">year</span> = 2021,
|
||||
<span class="org-variable-name">month</span> = {Nov},
|
||||
}
|
||||
</pre>
|
||||
</div>
|
||||
|
||||
<p>
|
||||
You can also use the formatted citation below.
|
||||
</p>
|
||||
<blockquote>
|
||||
<p>
|
||||
Dehaeze, T., Vermat, M., & Collette, C., A new method of designing complementary filters for sensor fusion using the \(H_\infty\) synthesis, Mechanical Systems and Signal Processing, (2021).
|
||||
</p>
|
||||
</blockquote>
|
||||
</div>
|
||||
</div>
|
||||
|
||||
<div id="outline-container-orgb8fb429" class="outline-2">
|
||||
<h2 id="orgb8fb429">Matlab Scripts (<a href="matlab/dehaeze21_desig_compl_filte_matlab.html">link</a>)</h2>
|
||||
<div class="outline-text-2" id="text-orgb8fb429">
|
||||
<div id="outline-container-org71c1b40" class="outline-2">
|
||||
<h2 id="org71c1b40">Matlab Scripts (<a href="matlab/dehaeze21_desig_compl_filte_matlab.html">link</a>)</h2>
|
||||
<div class="outline-text-2" id="text-org71c1b40">
|
||||
<p>
|
||||
All the <a href="https://fr.mathworks.com/">Matlab</a> code that was used for the paper are accessible so that all the results are reproducible.
|
||||
The source code is hosted on <a href="https://zenodo.org/record/5361920">Zenodo</a>.
|
||||
</p>
|
||||
|
||||
<p>
|
||||
If you use this code and want to refer to it in a publication, you can use the following BibTeX reference:
|
||||
</p>
|
||||
<div class="org-src-container">
|
||||
<pre class="src src-bibtex"><span class="org-function-name">@misc</span>{<span class="org-constant">dehaeze21_new_method_desig_compl_filter_code</span>,
|
||||
<span class="org-variable-name">author</span> = {Dehaeze, Thomas},
|
||||
<span class="org-variable-name">doi</span> = {<span class="org-button">10.5281/zenodo.5361943</span>},
|
||||
<span class="org-variable-name">howpublished</span> = {Source code on Zenodo},
|
||||
<span class="org-variable-name">month</span> = sep,
|
||||
<span class="org-variable-name">title</span> = {{A New Method of Designing Complementary Filters for Sensor
|
||||
Fusion Using the <span class="org-string">$\mathcal{H}_\infty$</span> Synthesis}},
|
||||
<span class="org-variable-name">year</span> = 2021,
|
||||
}
|
||||
</pre>
|
||||
</div>
|
||||
|
||||
<p>
|
||||
You can also use the formatted citation below.
|
||||
</p>
|
||||
<blockquote>
|
||||
<p>
|
||||
Dehaeze Thomas. (2021). A New Method of Designing Complementary Filters for Sensor Fusion Using the H-Infinity Synthesis. Source code on Zenodo. 10.5281/zenodo.5361943
|
||||
</p>
|
||||
</blockquote>
|
||||
</div>
|
||||
</div>
|
||||
|
||||
<div id="outline-container-org0ecd5c8" class="outline-2">
|
||||
<h2 id="org0ecd5c8">Tikz Figures (<a href="tikz/dehaeze21_desig_compl_filte_tikz.html">link</a>)</h2>
|
||||
<div class="outline-text-2" id="text-org0ecd5c8">
|
||||
<div id="outline-container-org1b73cc9" class="outline-2">
|
||||
<h2 id="org1b73cc9">Tikz Figures (<a href="tikz/dehaeze21_desig_compl_filte_tikz.html">link</a>)</h2>
|
||||
<div class="outline-text-2" id="text-org1b73cc9">
|
||||
<p>
|
||||
All the figures for the paper have been generated using <a href="https://sourceforge.net/projects/pgf/">TikZ</a>.
|
||||
</p>
|
||||
|
||||
68
index.org
68
index.org
@@ -1,4 +1,4 @@
|
||||
#+TITLE: Complementary Filters Shaping Using $\mathcal{H}_\infty$ Synthesis
|
||||
#+TITLE: A new method of designing complementary filters for sensor fusion using the $\mathcal{H}_\infty$ synthesis
|
||||
:DRAWER:
|
||||
#+SUBTITLE: Dehaeze Thomas, Verma Mohit and Collette Christophe
|
||||
|
||||
@@ -15,37 +15,69 @@
|
||||
#+begin_quote
|
||||
*Abstract*:
|
||||
|
||||
For many applications, large bandwidth and dynamic ranges are requiring to use several sensors, whose signals are combined using complementary filters.
|
||||
This paper presents a method for designing these complementary filters using $\mathcal{H}_\infty$ synthesis that allows to shape the filter norms.
|
||||
This method is shown to be easily applicable for the synthesis of complex complementary filters.
|
||||
In order to obtain a better estimate of a quantity being measured, several sensors having different characteristics can be merged with a technique called "sensor fusion".
|
||||
The obtained "super sensor" can combine the benefits of the individual sensors provided that the complementary filters used in the fusion are well designed.
|
||||
Indeed, properties of the super sensor are linked to the magnitude of the complementary filters.
|
||||
Properly shaping the magnitude of complementary filters is a difficult and time-consuming task.
|
||||
In this study, we address this issue and propose a new method for designing complementary filters.
|
||||
This method uses weighting functions to specify the wanted shape of the complementary filter that are then easily obtained using the standard $\mathcal{H}_\infty$ synthesis.
|
||||
The proper choice of the weighting functions is discussed, and the effectiveness and simplicity of the design method is highlighted using several examples.
|
||||
Such synthesis method is further extended for the shaping of more than two complementary filters.
|
||||
#+end_quote
|
||||
|
||||
* Paper ([[file:journal/dehaeze21_desig_compl_filte.pdf][link]])
|
||||
* Journal Article ([[file:journal/dehaeze21_desig_compl_filte.pdf][link]])
|
||||
:PROPERTIES:
|
||||
:UNNUMBERED: t
|
||||
:END:
|
||||
|
||||
The paper has been created [[https://orgmode.org/][Org Mode]] (generating [[https://www.latex-project.org/][LaTeX]] code) under [[https://www.gnu.org/software/emacs/][Emacs]].
|
||||
|
||||
To cite this paper use the following BibTeX code.
|
||||
#+begin_src bibtex
|
||||
@article{dehaeze21_new_method_desig_compl_filter,
|
||||
author = {Dehaeze, Thomas and Vermat, Mohit and Collette, Christophe},
|
||||
title = {A New Method of Designing Complementary Filters for Sensor
|
||||
Fusion Using the $\mathcal{H}_\infty$ Synthesis},
|
||||
journal = {Mechanical Systems and Signal Processing},
|
||||
year = 2021,
|
||||
month = {Nov},
|
||||
}
|
||||
#+end_src
|
||||
|
||||
You can also use the formatted citation below.
|
||||
#+begin_quote
|
||||
Dehaeze, T., Vermat, M., & Collette, C., A new method of designing complementary filters for sensor fusion using the $H_\infty$ synthesis, Mechanical Systems and Signal Processing, (2021).
|
||||
#+end_quote
|
||||
|
||||
* Matlab Scripts ([[file:matlab/dehaeze21_desig_compl_filte_matlab.org][link]])
|
||||
:PROPERTIES:
|
||||
:UNNUMBERED: t
|
||||
:END:
|
||||
All the [[https://fr.mathworks.com/][Matlab]] code that was used for the paper are accessible so that all the results are reproducible.
|
||||
|
||||
* Tikz Figures ([[file:tikz/index.org][link]])
|
||||
All the [[https://fr.mathworks.com/][Matlab]] code that was used for the paper are accessible so that all the results are reproducible.
|
||||
The source code is hosted on [[https://zenodo.org/record/5361920][Zenodo]].
|
||||
|
||||
If you use this code and want to refer to it in a publication, you can use the following BibTeX reference:
|
||||
#+begin_src bibtex
|
||||
@misc{dehaeze21_new_method_desig_compl_filter_code,
|
||||
author = {Dehaeze, Thomas},
|
||||
doi = {10.5281/zenodo.5361943},
|
||||
howpublished = {Source code on Zenodo},
|
||||
month = sep,
|
||||
title = {{A New Method of Designing Complementary Filters for Sensor
|
||||
Fusion Using the $\mathcal{H}_\infty$ Synthesis}},
|
||||
year = 2021,
|
||||
}
|
||||
#+end_src
|
||||
|
||||
You can also use the formatted citation below.
|
||||
#+begin_quote
|
||||
Dehaeze Thomas. (2021). A New Method of Designing Complementary Filters for Sensor Fusion Using the H-Infinity Synthesis. Source code on Zenodo. 10.5281/zenodo.5361943
|
||||
#+end_quote
|
||||
|
||||
* Tikz Figures ([[file:tikz/dehaeze21_desig_compl_filte_tikz.org][link]])
|
||||
:PROPERTIES:
|
||||
:UNNUMBERED: t
|
||||
:END:
|
||||
All the figures for the paper have been generated using [[https://sourceforge.net/projects/pgf/][TikZ]].
|
||||
|
||||
* Cite this paper :noexport:
|
||||
:PROPERTIES:
|
||||
:UNNUMBERED: t
|
||||
:END:
|
||||
To cite this paper use the following bibtex code.
|
||||
#+begin_src bibtex
|
||||
#+end_src
|
||||
|
||||
You can also use the formatted citation below.
|
||||
#+begin_quote
|
||||
#+end_quote
|
||||
|
||||
@@ -1,4 +1,4 @@
|
||||
#+TITLE: A new method of designing complementary filters for sensor fusion using the $\mathcal{H}_\infty$ synthesis
|
||||
#+TITLE: Designing complementary filters for sensor fusion using $\mathcal{H}_\infty$ synthesis
|
||||
:DRAWER:
|
||||
#+LATEX_CLASS: elsarticle
|
||||
#+LATEX_CLASS_OPTIONS: [preprint, sort&compress]
|
||||
@@ -12,8 +12,8 @@
|
||||
#+LATEX_HEADER_EXTRA: \author[a2,a4]{Christophe Collette}
|
||||
#+LATEX_HEADER_EXTRA: \cortext[cor1]{Corresponding author. Email Address: thomas.dehaeze@esrf.fr}
|
||||
|
||||
#+LATEX_HEADER_EXTRA: \address[a1]{European Synchrotron Radiation Facility, Grenoble, France}
|
||||
#+LATEX_HEADER_EXTRA: \address[a2]{University of Li\`{e}ge, Department of Aerospace and Mechanical Engineering, 4000 Li\`{e}ge, Belgium.}
|
||||
#+LATEX_HEADER_EXTRA: \address[a1]{European Synchrotron Radiation Facility, 38000 Grenoble, France}
|
||||
#+LATEX_HEADER_EXTRA: \address[a2]{University of Li\`{e}ge, PML, Department of Aerospace and Mechanical Engineering, 4000 Li\`{e}ge, Belgium.}
|
||||
#+LATEX_HEADER_EXTRA: \address[a3]{CSIR --- Structural Engineering Research Centre, Taramani, Chennai --- 600113, India.}
|
||||
#+LATEX_HEADER_EXTRA: \address[a4]{Universit\'{e} Libre de Bruxelles, Precision Mechatronics Laboratory, BEAMS Department, 1050 Brussels, Belgium.}
|
||||
|
||||
@@ -41,7 +41,16 @@
|
||||
("\\paragraph{%s}" . "\\paragraph*{%s}")
|
||||
("\\subparagraph{%s}" . "\\subparagraph*{%s}"))
|
||||
)
|
||||
;; Remove automatic org headings
|
||||
(defun my-latex-filter-removeOrgAutoLabels (text backend info)
|
||||
"Org-mode automatically generates labels for headings despite explicit use of `#+LABEL`. This filter forcibly removes all automatically generated org-labels in headings."
|
||||
(when (org-export-derived-backend-p backend 'latex)
|
||||
(replace-regexp-in-string "\\\\label{sec:org[a-f0-9]+}\n" "" text)))
|
||||
|
||||
(add-to-list 'org-export-filter-headline-functions
|
||||
'my-latex-filter-removeOrgAutoLabels)
|
||||
|
||||
;; Automatic delete org org-comments
|
||||
(defun delete-org-comments (backend)
|
||||
(loop for comment in (reverse (org-element-map (org-element-parse-buffer)
|
||||
'comment 'identity))
|
||||
@@ -65,13 +74,13 @@
|
||||
#+begin_frontmatter
|
||||
#+begin_abstract
|
||||
In order to obtain a better estimate of a quantity being measured, several sensors having different characteristics can be merged with a technique called "sensor fusion".
|
||||
The obtained "super sensor" can combine the benefits of the individual sensors provided that the complementary filters used in the fusion are well designed.
|
||||
The obtained "super sensor" combines the benefits of the individual sensors provided that the complementary filters used in the fusion are well designed.
|
||||
Indeed, properties of the super sensor are linked to the magnitude of the complementary filters.
|
||||
Properly shaping the magnitude of complementary filters is a difficult and time-consuming task.
|
||||
In this study, we address this issue and propose a new method for designing complementary filters.
|
||||
This method uses weighting functions to specify the wanted shape of the complementary filter that are then easily obtained using the standard $\mathcal{H}_\infty$ synthesis.
|
||||
In this study, this issue is addressed and a new method for designing complementary filters is proposed.
|
||||
This method uses weighting functions to specify the wanted shape of the complementary filters that are then obtained using the standard $\mathcal{H}_\infty$ synthesis.
|
||||
The proper choice of the weighting functions is discussed, and the effectiveness and simplicity of the design method is highlighted using several examples.
|
||||
Such synthesis method is further extended for the shaping of more than two complementary filters.
|
||||
Such synthesis method is further extended for the shaping of a set of more than two complementary filters.
|
||||
#+end_abstract
|
||||
|
||||
#+begin_keyword
|
||||
@@ -84,75 +93,75 @@ Sensor fusion \sep{} Complementary filters \sep{} $\mathcal{H}_\infty$ synthesis
|
||||
** Introduction to Sensor Fusion :ignore:
|
||||
|
||||
Measuring a physical quantity using sensors is always subject to several limitations.
|
||||
First, the accuracy of the measurement will be affected by several noise sources, such as the electrical noise of the conditioning electronics being used.
|
||||
First, the accuracy of the measurement is affected by several noise sources, such as electrical noise of the conditioning electronics being used.
|
||||
Second, the frequency range in which the measurement is relevant is bounded by the bandwidth of the sensor.
|
||||
One way to overcome these limitations is to combine several sensors using a technique called "sensor fusion"\nbsp{}cite:bendat57_optim_filter_indep_measur_two.
|
||||
Fortunately, a wide variety of sensors exist, each with different characteristics.
|
||||
By carefully choosing the fused sensors, a so called "super sensor" is obtain that combines benefits of individual sensors and yields a better estimate of the measured physical quantity. \par
|
||||
Fortunately, a wide variety of sensors exists, each with different characteristics.
|
||||
By carefully choosing the fused sensors, a so called "super sensor" is obtained that can combines benefits of the individual sensors. \par
|
||||
|
||||
** Advantages of Sensor Fusion :ignore:
|
||||
|
||||
In some situations, sensor fusion is used to increase the bandwidth of the measurement\nbsp{}cite:shaw90_bandw_enhan_posit_measur_using_measur_accel,zimmermann92_high_bandw_orien_measur_contr,min15_compl_filter_desig_angle_estim.
|
||||
For instance, in\nbsp{}cite:shaw90_bandw_enhan_posit_measur_using_measur_accel, the bandwidth of a position sensor is increased by fusing it with an accelerometer providing the high frequency motion information.
|
||||
For other applications, sensor fusion is used to obtain a estimate of the measured quantity with lower noise\nbsp{}cite:hua05_low_ligo,hua04_polyp_fir_compl_filter_contr_system,plummer06_optim_compl_filter_their_applic_motion_measur,robert12_introd_random_signal_applied_kalman.
|
||||
More recently, the fusion of sensors measuring different physical quantities has been proposed to obtained interesting properties for control\nbsp{}cite:collette15_sensor_fusion_method_high_perfor,yong16_high_speed_vertic_posit_stage.
|
||||
In\nbsp{}cite:collette15_sensor_fusion_method_high_perfor, an inertial sensor used for active vibration isolation is fused with a sensor collocated with the actuator to improve the stability margins of the feedback controller. \par
|
||||
For instance, in\nbsp{}cite:shaw90_bandw_enhan_posit_measur_using_measur_accel the bandwidth of a position sensor is increased by fusing it with an accelerometer providing the high frequency motion information.
|
||||
For other applications, sensor fusion is used to obtain an estimate of the measured quantity with lower noise\nbsp{}cite:hua05_low_ligo,hua04_polyp_fir_compl_filter_contr_system,plummer06_optim_compl_filter_their_applic_motion_measur,robert12_introd_random_signal_applied_kalman.
|
||||
More recently, the fusion of sensors measuring different physical quantities has been proposed to obtain interesting properties for control\nbsp{}cite:collette15_sensor_fusion_method_high_perfor,yong16_high_speed_vertic_posit_stage.
|
||||
In\nbsp{}cite:collette15_sensor_fusion_method_high_perfor, an inertial sensor used for active vibration isolation is fused with a sensor collocated with the actuator for improving the stability margins of the feedback controller. \par
|
||||
|
||||
** Applications :ignore:
|
||||
|
||||
Practical applications of sensor fusion are numerous.
|
||||
It is widely used for the attitude estimation of several autonomous vehicles such as unmanned aerial vehicle\nbsp{}cite:baerveldt97_low_cost_low_weigh_attit,corke04_inert_visual_sensin_system_small_auton_helic,jensen13_basic_uas and underwater vehicles\nbsp{}cite:pascoal99_navig_system_desig_using_time,batista10_optim_posit_veloc_navig_filter_auton_vehic.
|
||||
Naturally, it is of great benefits for high performance positioning control as shown in\nbsp{}cite:shaw90_bandw_enhan_posit_measur_using_measur_accel,zimmermann92_high_bandw_orien_measur_contr,min15_compl_filter_desig_angle_estim,yong16_high_speed_vertic_posit_stage.
|
||||
Sensor fusion was also shown to be a key technology to improve the performances of active vibration isolation systems\nbsp{}cite:tjepkema12_sensor_fusion_activ_vibrat_isolat_precis_equip.
|
||||
This is particularly apparent for the isolation stages of gravitational wave observer\nbsp{}cite:collette15_sensor_fusion_method_high_perfor,heijningen18_low such as the ones used at the LIGO\nbsp{}cite:hua05_low_ligo,hua04_polyp_fir_compl_filter_contr_system and at the VIRGO\nbsp{}cite:lucia18_low_frequen_optim_perfor_advan. \par
|
||||
Sensor fusion was also shown to be a key technology to improve the performance of active vibration isolation systems\nbsp{}cite:tjepkema12_sensor_fusion_activ_vibrat_isolat_precis_equip.
|
||||
Emblematic examples are the isolation stages of gravitational wave detectors\nbsp{}cite:collette15_sensor_fusion_method_high_perfor,heijningen18_low such as the ones used at the LIGO\nbsp{}cite:hua05_low_ligo,hua04_polyp_fir_compl_filter_contr_system and at the Virgo\nbsp{}cite:lucia18_low_frequen_optim_perfor_advan. \par
|
||||
|
||||
** Kalman Filtering / Complementary filters :ignore:
|
||||
|
||||
There are mainly two ways to perform sensor fusion: either using a set of complementary filters\nbsp{}cite:anderson53_instr_approac_system_steer_comput or using Kalman filtering\nbsp{}cite:brown72_integ_navig_system_kalman_filter,odry18_kalman_filter_mobil_robot_attit_estim.
|
||||
For sensor fusion applications, both methods are sharing many relationships\nbsp{}cite:brown72_integ_navig_system_kalman_filter,higgins75_compar_compl_kalman_filter,robert12_introd_random_signal_applied_kalman,becker15_compl_filter_desig_three_frequen_bands.
|
||||
However, for Kalman filtering, we are forced to make assumption about the probabilistic character of the sensor noises\nbsp{}cite:robert12_introd_random_signal_applied_kalman whereas it is not the case for complementary filters.
|
||||
Furthermore, the main advantages of complementary filters over Kalman filtering for sensor fusion are their very general applicability, their low computational cost\nbsp{}cite:higgins75_compar_compl_kalman_filter, and the fact that they are very intuitive as their effects can be easily interpreted in the frequency domain. \par
|
||||
However, for Kalman filtering, assumptions must be made about the probabilistic character of the sensor noises\nbsp{}cite:robert12_introd_random_signal_applied_kalman whereas it is not the case with complementary filters.
|
||||
Furthermore, the advantages of complementary filters over Kalman filtering for sensor fusion are their general applicability, their low computational cost\nbsp{}cite:higgins75_compar_compl_kalman_filter, and the fact that they are intuitive as their effects can be easily interpreted in the frequency domain. \par
|
||||
|
||||
** Design Methods of Complementary filters :ignore:
|
||||
|
||||
A set of filters is said to be complementary if the sum of their transfer functions is equal to one at all frequencies.
|
||||
For the earliest used of complementary filtering for sensor fusion, analog circuits were used to physically realize the filters\nbsp{}cite:anderson53_instr_approac_system_steer_comput.
|
||||
Analog complementary filters are still used today\nbsp{}cite:yong16_high_speed_vertic_posit_stage,moore19_capac_instr_sensor_fusion_high_bandw_nanop, but most of the time they are now implemented numerically as it allows for much more flexibility. \par
|
||||
In the early days of complementary filtering, analog circuits were employed to physically realize the filters\nbsp{}cite:anderson53_instr_approac_system_steer_comput.
|
||||
Analog complementary filters are still used today\nbsp{}cite:yong16_high_speed_vertic_posit_stage,moore19_capac_instr_sensor_fusion_high_bandw_nanop, but most of the time they are now implemented digitally as it allows for much more flexibility. \par
|
||||
|
||||
Several design methods have been developed over the years to optimize complementary filters.
|
||||
The easiest way to design complementary filters is to use analytical formulas.
|
||||
Depending on the application, the formulas used are of first order\nbsp{}cite:corke04_inert_visual_sensin_system_small_auton_helic,yeh05_model_contr_hydraul_actuat_two,yong16_high_speed_vertic_posit_stage, second order\nbsp{}cite:baerveldt97_low_cost_low_weigh_attit,stoten01_fusion_kinet_data_using_compos_filter,jensen13_basic_uas or even higher orders\nbsp{}cite:shaw90_bandw_enhan_posit_measur_using_measur_accel,zimmermann92_high_bandw_orien_measur_contr,stoten01_fusion_kinet_data_using_compos_filter,collette15_sensor_fusion_method_high_perfor,matichard15_seism_isolat_advan_ligo. \par
|
||||
|
||||
As the characteristics of the "super sensor" depends on the design of the complementary filters\nbsp{}cite:dehaeze19_compl_filter_shapin_using_synth, several optimization techniques were developed over the years.
|
||||
Some are based on the finding the optimal parameters of analytical formulas\nbsp{}cite:jensen13_basic_uas,min15_compl_filter_desig_angle_estim,becker15_compl_filter_desig_three_frequen_bands, while other are using convex optimization tools\nbsp{}cite:hua04_polyp_fir_compl_filter_contr_system,hua05_low_ligo such as linear matrix inequalities\nbsp{}cite:pascoal99_navig_system_desig_using_time.
|
||||
As the characteristics of the super sensor depends on the proper design of the complementary filters\nbsp{}cite:dehaeze19_compl_filter_shapin_using_synth, several optimization techniques have been developed.
|
||||
Some are based on the finding of optimal parameters of analytical formulas\nbsp{}cite:jensen13_basic_uas,min15_compl_filter_desig_angle_estim,becker15_compl_filter_desig_three_frequen_bands, while other are using convex optimization tools\nbsp{}cite:hua04_polyp_fir_compl_filter_contr_system,hua05_low_ligo such as linear matrix inequalities\nbsp{}cite:pascoal99_navig_system_desig_using_time.
|
||||
As shown in\nbsp{}cite:plummer06_optim_compl_filter_their_applic_motion_measur, the design of complementary filters can also be linked to the standard mixed-sensitivity control problem.
|
||||
Therefore, all the powerful tools developed for the classical control theory can also be used for the design of complementary filters.
|
||||
For instance, in\nbsp{}cite:jensen13_basic_uas the two gains of a Proportional Integral (PI) controller are optimized to minimize the noise of the super sensor. \par
|
||||
|
||||
** Problematic / gap in the research :ignore:
|
||||
|
||||
The common objective to all these complementary filters design methods is to obtain a super sensor that has desired characteristics, usually in terms of noise and dynamics.
|
||||
Moreover, as reported in\nbsp{}cite:zimmermann92_high_bandw_orien_measur_contr,plummer06_optim_compl_filter_their_applic_motion_measur, phase shifts and magnitude bumps of the "super sensors" dynamics can be observed if either the complementary filters are poorly designed or if the sensors are not well calibrated.
|
||||
Therefore, the robustness of the fusion is also of concerned when designing the complementary filters.
|
||||
Although many design methods of complementary filters have been proposed in the literature, no simple method that allows to specify the wanted super sensor characteristic while ensuring good fusion robustness has been proposed. \par
|
||||
The common objective of all these complementary filters design methods is to obtain a super sensor that has desired characteristics, usually in terms of noise and dynamics.
|
||||
Moreover, as reported in\nbsp{}cite:zimmermann92_high_bandw_orien_measur_contr,plummer06_optim_compl_filter_their_applic_motion_measur, phase shifts and magnitude bumps of the super sensors dynamics can be observed if either the complementary filters are poorly designed or if the sensors are not well calibrated.
|
||||
Hence, the robustness of the fusion is also of concern when designing the complementary filters.
|
||||
Although many design methods of complementary filters have been proposed in the literature, no simple method that allows to specify the desired super sensor characteristic while ensuring good fusion robustness has been proposed. \par
|
||||
|
||||
** Describe the paper itself / the problem which is addressed :ignore:
|
||||
|
||||
Fortunately, both the robustness of the fusion and the super sensor characteristics can be linked to the magnitude of the complementary filters\nbsp{}cite:dehaeze19_compl_filter_shapin_using_synth.
|
||||
Based on that, this paper introduces a new way to design complementary filters using the $\mathcal{H}_\infty$ synthesis which allows to shape the magnitude of the complementary filters in an easy and intuitive way. \par
|
||||
Based on that, this paper introduces a new way to design complementary filters using the $\mathcal{H}_\infty$ synthesis which allows to shape the complementary filters' magnitude in an easy and intuitive way. \par
|
||||
|
||||
** Introduce Each part of the paper :ignore:
|
||||
|
||||
Section\nbsp{}ref:sec:requirements introduces the sensor fusion architecture and demonstrates how typical requirements can be linked to the complementary filters magnitudes.
|
||||
In section\nbsp{}ref:sec:hinf_method, the shaping of complementary filters is written as an $\mathcal{H}_\infty$ optimization problem using weighting functions, and the simplicity of the proposed method is illustrated with an example.
|
||||
Section\nbsp{}ref:sec:requirements introduces the sensor fusion architecture and demonstrates how typical requirements can be linked to the complementary filters' magnitude.
|
||||
In Section\nbsp{}ref:sec:hinf_method, the shaping of complementary filters is formulated as an $\mathcal{H}_\infty$ optimization problem using weighting functions, and the simplicity of the proposed method is illustrated with an example.
|
||||
The synthesis method is further validated in Section\nbsp{}ref:sec:application_ligo by designing complex complementary filters.
|
||||
Section\nbsp{}ref:sec:discussion compares the proposed synthesis method with the classical mixed-sensitivity synthesis, and extends it to the shaping of more than two complementary filters.
|
||||
Section\nbsp{}ref:sec:discussion compares the proposed synthesis method with the classical mixed-sensitivity synthesis, and extends it for the shaping of more than two complementary filters.
|
||||
|
||||
* Sensor Fusion and Complementary Filters Requirements
|
||||
<<sec:requirements>>
|
||||
** Introduction :ignore:
|
||||
|
||||
Complementary filters provides a framework for fusing signals from different sensors.
|
||||
Complementary filtering provides a framework for fusing signals from different sensors.
|
||||
As the effectiveness of the fusion depends on the proper design of the complementary filters, they are expected to fulfill certain requirements.
|
||||
These requirements are discussed in this section.
|
||||
|
||||
@@ -161,16 +170,16 @@ These requirements are discussed in this section.
|
||||
|
||||
A general sensor fusion architecture using complementary filters is shown in Fig.\nbsp{}ref:fig:sensor_fusion_overview where several sensors (here two) are measuring the same physical quantity $x$.
|
||||
The two sensors output signals $\hat{x}_1$ and $\hat{x}_2$ are estimates of $x$.
|
||||
Each of these estimates are then filtered out by complementary filters and combined to form a new estimate $\hat{x}$.
|
||||
These estimates are then filtered out by complementary filters and combined to form a new estimate $\hat{x}$.
|
||||
|
||||
The resulting sensor, termed as super sensor, can have larger bandwidth and better noise characteristics in comparison to the individual sensor.
|
||||
The resulting sensor, termed as super sensor, can have larger bandwidth and better noise characteristics in comparison to the individual sensors.
|
||||
This means that the super sensor provides an estimate $\hat{x}$ of $x$ which can be more accurate over a larger frequency band than the outputs of the individual sensors.
|
||||
|
||||
#+name: fig:sensor_fusion_overview
|
||||
#+caption: Schematic of a sensor fusion architecture using complementary filters
|
||||
#+caption: Schematic of a sensor fusion architecture using complementary filters.
|
||||
[[file:figs/sensor_fusion_overview.pdf]]
|
||||
|
||||
The complementary property of filters $H_1(s)$ and $H_2(s)$ implies that the sum of their transfer functions is equal to unity.
|
||||
The complementary property of filters $H_1(s)$ and $H_2(s)$ implies that the sum of their transfer functions is equal to one.
|
||||
That is, unity magnitude and zero phase at all frequencies.
|
||||
Therefore, a pair of complementary filter needs to satisfy the following condition:
|
||||
#+name: eq:comp_filter
|
||||
@@ -191,22 +200,22 @@ Before filtering the sensor outputs $\tilde{x}_i$ by the complementary filters,
|
||||
This normalization consists of using an estimate $\hat{G}_i(s)$ of the sensor dynamics $G_i(s)$, and filtering the sensor output by the inverse of this estimate $\hat{G}_i^{-1}(s)$ as shown in Fig.\nbsp{}ref:fig:sensor_model_calibrated.
|
||||
It is here supposed that the sensor inverse $\hat{G}_i^{-1}(s)$ is proper and stable.
|
||||
This way, the units of the estimates $\hat{x}_i$ are equal to the units of the physical quantity $x$.
|
||||
The sensor dynamics estimate $\hat{G}_1(s)$ can be a simple gain or a more complex transfer function.
|
||||
The sensor dynamics estimate $\hat{G}_i(s)$ can be a simple gain or a more complex transfer function.
|
||||
|
||||
#+begin_export latex
|
||||
\begin{figure}[htbp]
|
||||
\begin{subfigure}[b]{0.49\linewidth}
|
||||
\centering
|
||||
\includegraphics[scale=1]{figs/sensor_model.pdf}
|
||||
\caption{\label{fig:sensor_model} Basic sensor model consisting of a noise input $n_i$ and a linear time invariant transfer function $G_i(s)$}
|
||||
\caption{\label{fig:sensor_model} Basic sensor model consisting of a noise input $n_i$ and a linear time invariant transfer function $G_i(s)$.}
|
||||
\end{subfigure}
|
||||
\hfill
|
||||
\begin{subfigure}[b]{0.49\linewidth}
|
||||
\centering
|
||||
\includegraphics[scale=1]{figs/sensor_model_calibrated.pdf}
|
||||
\caption{\label{fig:sensor_model_calibrated} Normalized sensors using the inverse of an estimate $\hat{G}_i(s)$ of the sensor dynamics}
|
||||
\caption{\label{fig:sensor_model_calibrated} Normalized sensors using the inverse of an estimate $\hat{G}_i(s)$ of the sensor dynamics.}
|
||||
\end{subfigure}
|
||||
\caption{\label{fig:sensor_models}Sensor models with and without normalization}
|
||||
\caption{\label{fig:sensor_models}Sensor models with and without normalization.}
|
||||
\centering
|
||||
\end{figure}
|
||||
#+end_export
|
||||
@@ -222,7 +231,7 @@ The super sensor output is therefore equal to:
|
||||
\end{equation}
|
||||
|
||||
#+name: fig:fusion_super_sensor
|
||||
#+caption: Sensor fusion architecture with two normalized sensors
|
||||
#+caption: Sensor fusion architecture with two normalized sensors.
|
||||
[[file:figs/fusion_super_sensor.pdf]]
|
||||
|
||||
** Noise Sensor Filtering
|
||||
@@ -236,23 +245,23 @@ In this section, it is supposed that all the sensors are perfectly normalized, s
|
||||
|
||||
The effect of a non-perfect normalization will be discussed in the next section.
|
||||
|
||||
Provided eqref:eq:perfect_dynamics is verified, the super sensor output $\hat{x}$ is then:
|
||||
Provided\nbsp{}eqref:eq:perfect_dynamics is verified, the super sensor output $\hat{x}$ is then equal to:
|
||||
#+name: eq:estimate_perfect_dyn
|
||||
\begin{equation}
|
||||
\hat{x} = x + H_1(s) n_1 + H_2(s) n_2
|
||||
\end{equation}
|
||||
|
||||
From eqref:eq:estimate_perfect_dyn, the complementary filters $H_1(s)$ and $H_2(s)$ are shown to only operate on the sensor's noises.
|
||||
From\nbsp{}eqref:eq:estimate_perfect_dyn, the complementary filters $H_1(s)$ and $H_2(s)$ are shown to only operate on the noise of the sensors.
|
||||
Thus, this sensor fusion architecture permits to filter the noise of both sensors without introducing any distortion in the physical quantity to be measured.
|
||||
This is why the two filters must be complementary.
|
||||
|
||||
The estimation error $\delta x$, defined as the difference between the sensor output $\hat{x}$ and the measured quantity $x$, is computed for the super sensor eqref:eq:estimate_error.
|
||||
The estimation error $\delta x$, defined as the difference between the sensor output $\hat{x}$ and the measured quantity $x$, is computed for the super sensor\nbsp{}eqref:eq:estimate_error.
|
||||
#+name: eq:estimate_error
|
||||
\begin{equation}
|
||||
\delta x \triangleq \hat{x} - x = H_1(s) n_1 + H_2(s) n_2
|
||||
\end{equation}
|
||||
|
||||
As shown in eqref:eq:noise_filtering_psd, the Power Spectral Density (PSD) of the estimation error $\Phi_{\delta x}$ depends both on the norm of the two complementary filters and on the PSD of the noise sources $\Phi_{n_1}$ and $\Phi_{n_2}$.
|
||||
As shown in\nbsp{}eqref:eq:noise_filtering_psd, the Power Spectral Density (PSD) of the estimation error $\Phi_{\delta x}$ depends both on the norm of the two complementary filters and on the PSD of the noise sources $\Phi_{n_1}$ and $\Phi_{n_2}$.
|
||||
#+name: eq:noise_filtering_psd
|
||||
\begin{equation}
|
||||
\Phi_{\delta x}(\omega) = \left|H_1(j\omega)\right|^2 \Phi_{n_1}(\omega) + \left|H_2(j\omega)\right|^2 \Phi_{n_2}(\omega)
|
||||
@@ -263,17 +272,17 @@ This is the simplest form of sensor fusion with complementary filters.
|
||||
|
||||
However, the two sensors have usually high noise levels over distinct frequency regions.
|
||||
In such case, to lower the noise of the super sensor, the norm $|H_1(j\omega)|$ has to be small when $\Phi_{n_1}(\omega)$ is larger than $\Phi_{n_2}(\omega)$ and the norm $|H_2(j\omega)|$ has to be small when $\Phi_{n_2}(\omega)$ is larger than $\Phi_{n_1}(\omega)$.
|
||||
Therefore, by properly shaping the norm of the complementary filters, it is possible to minimize the noise of the super sensor noise.
|
||||
Hence, by properly shaping the norm of the complementary filters, it is possible to reduce the noise of the super sensor.
|
||||
|
||||
** Sensor Fusion Robustness
|
||||
<<sec:fusion_robustness>>
|
||||
|
||||
In practical systems the sensor normalization is not perfect and condition eqref:eq:perfect_dynamics is not verified.
|
||||
In practical systems the sensor normalization is not perfect and condition\nbsp{}eqref:eq:perfect_dynamics is not verified.
|
||||
|
||||
In order to study such imperfection, a multiplicative input uncertainty is added to the sensor dynamics (Fig.\nbsp{}ref:fig:sensor_model_uncertainty).
|
||||
The nominal model is taken as the estimated model for the normalization $\hat{G}_i(s)$, $\Delta_i$ is any stable transfer function satisfying $|\Delta_i(j\omega)| \le 1,\ \forall\omega$, and $w_i(s)$ is a weighting transfer function representing the magnitude of the uncertainty.
|
||||
The nominal model is the estimated model used for the normalization $\hat{G}_i(s)$, $\Delta_i(s)$ is any stable transfer function satisfying $|\Delta_i(j\omega)| \le 1,\ \forall\omega$, and $w_i(s)$ is a weighting transfer function representing the magnitude of the uncertainty.
|
||||
|
||||
The weight $w_i(s)$ is chosen such that the real sensor dynamics $G(j\omega)$ is contained in the uncertain region represented by a circle in the complex plane, centered on $1$ and with a radius equal to $|w_i(j\omega)|$.
|
||||
The weight $w_i(s)$ is chosen such that the real sensor dynamics $G_i(j\omega)$ is contained in the uncertain region represented by a circle in the complex plane, centered on $1$ and with a radius equal to $|w_i(j\omega)|$.
|
||||
|
||||
As the nominal sensor dynamics is taken as the normalized filter, the normalized sensor can be further simplified as shown in Fig.\nbsp{}ref:fig:sensor_model_uncertainty_simplified.
|
||||
|
||||
@@ -282,41 +291,40 @@ As the nominal sensor dynamics is taken as the normalized filter, the normalized
|
||||
\begin{subfigure}[b]{0.59\linewidth}
|
||||
\centering
|
||||
\includegraphics[scale=1]{figs/sensor_model_uncertainty.pdf}
|
||||
\caption{\label{fig:sensor_model_uncertainty} Sensor with multiplicative input uncertainty}
|
||||
\caption{\label{fig:sensor_model_uncertainty} Sensor with multiplicative input uncertainty.}
|
||||
\end{subfigure}
|
||||
\hfill
|
||||
\begin{subfigure}[b]{0.39\linewidth}
|
||||
\centering
|
||||
\includegraphics[scale=1]{figs/sensor_model_uncertainty_simplified.pdf}
|
||||
\caption{\label{fig:sensor_model_uncertainty_simplified} Simplified sensor model}
|
||||
\caption{\label{fig:sensor_model_uncertainty_simplified} Simplified sensor model.}
|
||||
\end{subfigure}
|
||||
\hfill
|
||||
\caption{\label{fig:sensor_models_uncertainty}Sensor models with dynamical uncertainty}
|
||||
\caption{\label{fig:sensor_models_uncertainty}Sensor models with dynamical uncertainty.}
|
||||
\centering
|
||||
\end{figure}
|
||||
#+end_export
|
||||
|
||||
The sensor fusion architecture with two sensor models including dynamical uncertainty is shown in Fig.\nbsp{}ref:fig:sensor_fusion_dynamic_uncertainty.
|
||||
The sensor fusion architecture with the sensor models including dynamical uncertainty is shown in Fig.\nbsp{}ref:fig:sensor_fusion_dynamic_uncertainty.
|
||||
|
||||
#+name: fig:sensor_fusion_dynamic_uncertainty
|
||||
#+caption: Sensor fusion architecture with sensor dynamics uncertainty
|
||||
#+caption: Sensor fusion architecture with sensor dynamics uncertainty.
|
||||
[[file:figs/sensor_fusion_dynamic_uncertainty.pdf]]
|
||||
|
||||
The super sensor dynamics eqref:eq:super_sensor_dyn_uncertainty is no longer equal to $1$ and now depends on the sensor dynamical uncertainty weights $w_i(s)$ as well as on the complementary filters $H_i(s)$.
|
||||
The super sensor dynamics\nbsp{}eqref:eq:super_sensor_dyn_uncertainty is no longer equal to $1$ and now depends on the sensor dynamical uncertainty weights $w_i(s)$ as well as on the complementary filters $H_i(s)$.
|
||||
#+name: eq:super_sensor_dyn_uncertainty
|
||||
\begin{equation}
|
||||
\frac{\hat{x}}{x} = 1 + w_1(s) H_1(s) \Delta_1(s) + w_2(s) H_2(s) \Delta_2(s)
|
||||
\end{equation}
|
||||
|
||||
The dynamical uncertainty of the super sensor can be graphically represented in the complex plane by a circle centered on $1$ with a radius equal to $|w_1(j\omega) H_1(j\omega)| + |w_2(j\omega) H_2(j\omega)|$ as shown in Fig.\nbsp{}ref:fig:uncertainty_set_super_sensor.
|
||||
The dynamical uncertainty of the super sensor can be graphically represented in the complex plane by a circle centered on $1$ with a radius equal to $|w_1(j\omega) H_1(j\omega)| + |w_2(j\omega) H_2(j\omega)|$ (Fig.\nbsp{}ref:fig:uncertainty_set_super_sensor).
|
||||
|
||||
#+name: fig:uncertainty_set_super_sensor
|
||||
#+caption: Uncertainty region of the super sensor dynamics in the complex plane (grey circle). The contribution of both sensors 1 and 2 to the total uncertainty are represented respectively by a blue circle and a red circle. The frequency dependency $\omega$ is here omitted.
|
||||
[[file:figs/uncertainty_set_super_sensor.pdf]]
|
||||
|
||||
The super sensor dynamical uncertainty (i.e. the robustness of the fusion) clearly depends on the complementary filters norms.
|
||||
For instance, the phase $\Delta\phi(\omega)$ added by the super sensor dynamics at frequency $\omega$ is bounded by $\Delta\phi_{\text{max}}(\omega)$ which can be found by drawing a tangent from the origin to the uncertainty circle of the super sensor (Fig.\nbsp{}ref:fig:uncertainty_set_super_sensor).
|
||||
Therefore, the phase uncertainty of the super sensor dynamics depends on the Complementary filters norms eqref:eq:max_phase_uncertainty.
|
||||
The super sensor dynamical uncertainty, and hence the robustness of the fusion, clearly depends on the complementary filters' norm.
|
||||
For instance, the phase $\Delta\phi(\omega)$ added by the super sensor dynamics at frequency $\omega$ is bounded by $\Delta\phi_{\text{max}}(\omega)$ which can be found by drawing a tangent from the origin to the uncertainty circle of the super sensor (Fig.\nbsp{}ref:fig:uncertainty_set_super_sensor) and that is mathematically described by\nbsp{}eqref:eq:max_phase_uncertainty.
|
||||
|
||||
#+name: eq:max_phase_uncertainty
|
||||
\begin{equation}
|
||||
@@ -330,17 +338,17 @@ Typically, the norm of the complementary filter $|H_i(j\omega)|$ should be made
|
||||
<<sec:hinf_method>>
|
||||
** Introduction :ignore:
|
||||
|
||||
As shown in Section\nbsp{}ref:sec:requirements, the noise and robustness of the super sensor are a function of the complementary filters norms.
|
||||
Therefore, a complementary filters synthesis method that allows to shape their norms would be of great use.
|
||||
As shown in Section\nbsp{}ref:sec:requirements, the noise and robustness of the super sensor are a function of the complementary filters' norm.
|
||||
Therefore, a synthesis method of complementary filters that allows to shape their norm would be of great use.
|
||||
In this section, such synthesis is proposed by writing the synthesis objective as a standard $\mathcal{H}_\infty$ optimization problem.
|
||||
As weighting functions are used to represent the wanted complementary filters shapes during the synthesis, the proper design of weighting functions is discussed.
|
||||
As weighting functions are used to represent the wanted complementary filters' shape during the synthesis, their proper design is discussed.
|
||||
Finally, the synthesis method is validated on an simple example.
|
||||
|
||||
** Synthesis Objective
|
||||
<<sec:synthesis_objective>>
|
||||
|
||||
The synthesis objective is to shape the norm of two filters $H_1(s)$ and $H_2(s)$ while ensuring their complementary property eqref:eq:comp_filter.
|
||||
This is equivalent as to finding proper and stable transfer functions $H_1(s)$ and $H_2(s)$ such that conditions eqref:eq:hinf_cond_complementarity, eqref:eq:hinf_cond_h1 and eqref:eq:hinf_cond_h2 are satisfied.
|
||||
The synthesis objective is to shape the norm of two filters $H_1(s)$ and $H_2(s)$ while ensuring their complementary property\nbsp{}eqref:eq:comp_filter.
|
||||
This is equivalent as to finding proper and stable transfer functions $H_1(s)$ and $H_2(s)$ such that conditions\nbsp{}eqref:eq:hinf_cond_complementarity, eqref:eq:hinf_cond_h1 and\nbsp{}eqref:eq:hinf_cond_h2 are satisfied.
|
||||
|
||||
#+name: eq:comp_filter_problem_form
|
||||
\begin{subequations}
|
||||
@@ -350,14 +358,14 @@ This is equivalent as to finding proper and stable transfer functions $H_1(s)$ a
|
||||
& |H_2(j\omega)| \le \frac{1}{|W_2(j\omega)|} \quad \forall\omega \label{eq:hinf_cond_h2}
|
||||
\end{align}
|
||||
\end{subequations}
|
||||
$W_1(s)$ and $W_2(s)$ are two weighting transfer functions that are carefully chosen to specify the maximum wanted norms of the complementary filters during the synthesis.
|
||||
$W_1(s)$ and $W_2(s)$ are two weighting transfer functions that are carefully chosen to specify the maximum wanted norm of the complementary filters during the synthesis.
|
||||
|
||||
** Shaping of Complementary Filters using $\mathcal{H}_\infty$ synthesis
|
||||
<<sec:hinf_synthesis>>
|
||||
|
||||
In this section, it is shown that the synthesis objective can be easily expressed as a standard $\mathcal{H}_\infty$ optimization problem and therefore solved using convenient tools readily available.
|
||||
|
||||
Consider the generalized plant $P(s)$ shown in Fig.\nbsp{}ref:fig:h_infinity_robust_fusion_plant and mathematically described by eqref:eq:generalized_plant.
|
||||
Consider the generalized plant $P(s)$ shown in Fig.\nbsp{}ref:fig:h_infinity_robust_fusion_plant and mathematically described by\nbsp{}eqref:eq:generalized_plant.
|
||||
|
||||
#+name: eq:generalized_plant
|
||||
\begin{equation}
|
||||
@@ -369,28 +377,28 @@ Consider the generalized plant $P(s)$ shown in Fig.\nbsp{}ref:fig:h_infinity_rob
|
||||
\begin{subfigure}[t]{0.5\linewidth}
|
||||
\centering
|
||||
\includegraphics[scale=1]{figs/h_infinity_robust_fusion_plant.pdf}
|
||||
\caption{\label{fig:h_infinity_robust_fusion_plant} Generalized plant}
|
||||
\caption{\label{fig:h_infinity_robust_fusion_plant} Generalized plant.}
|
||||
\vfill
|
||||
\end{subfigure}
|
||||
\hfill
|
||||
\begin{subfigure}[t]{0.5\linewidth}
|
||||
\centering
|
||||
\includegraphics[scale=1]{figs/h_infinity_robust_fusion_fb.pdf}
|
||||
\caption{\label{fig:h_infinity_robust_fusion_fb} Generalized plant with the synthesized filter}
|
||||
\caption{\label{fig:h_infinity_robust_fusion_fb} Generalized plant with the synthesized filter.}
|
||||
\end{subfigure}
|
||||
\hfill
|
||||
\caption{\label{fig:h_infinity_robust_fusion} Architecture for the $\mathcal{H}_\infty$ synthesis of complementary filters}
|
||||
\caption{\label{fig:h_infinity_robust_fusion} Architecture for the $\mathcal{H}_\infty$ synthesis of complementary filters.}
|
||||
\centering
|
||||
\end{figure}
|
||||
#+end_export
|
||||
|
||||
Applying the standard $\mathcal{H}_\infty$ synthesis on the generalized plant $P(s)$ is then equivalent as finding a stable filter $H_2(s)$ which based on $v$, generates a signal $u$ such that the $\mathcal{H}_\infty$ norm of the system in Fig.\nbsp{}ref:fig:h_infinity_robust_fusion_fb from $w$ to $[z_1, \ z_2]$ is less than one eqref:eq:hinf_syn_obj.
|
||||
Applying the standard $\mathcal{H}_\infty$ synthesis to the generalized plant $P(s)$ is then equivalent as finding a stable filter $H_2(s)$ which based on $v$, generates a signal $u$ such that the $\mathcal{H}_\infty$ norm of the system in Fig.\nbsp{}ref:fig:h_infinity_robust_fusion_fb from $w$ to $[z_1, \ z_2]$ is less than one\nbsp{}eqref:eq:hinf_syn_obj.
|
||||
#+name: eq:hinf_syn_obj
|
||||
\begin{equation}
|
||||
\left\|\begin{matrix} \left(1 - H_2(s)\right) W_1(s) \\ H_2(s) W_2(s) \end{matrix}\right\|_\infty \le 1
|
||||
\end{equation}
|
||||
|
||||
By then defining $H_1(s)$ to be the complementary of $H_2(s)$ eqref:eq:definition_H1, the $\mathcal{H}_\infty$ synthesis objective becomes equivalent to eqref:eq:hinf_problem which ensure that eqref:eq:hinf_cond_h1 and eqref:eq:hinf_cond_h2 are satisfied.
|
||||
By then defining $H_1(s)$ to be the complementary of $H_2(s)$ eqref:eq:definition_H1, the $\mathcal{H}_\infty$ synthesis objective becomes equivalent to\nbsp{}eqref:eq:hinf_problem which ensures that\nbsp{}eqref:eq:hinf_cond_h1 and\nbsp{}eqref:eq:hinf_cond_h2 are satisfied.
|
||||
|
||||
#+name: eq:definition_H1
|
||||
\begin{equation}
|
||||
@@ -402,26 +410,25 @@ By then defining $H_1(s)$ to be the complementary of $H_2(s)$ eqref:eq:definitio
|
||||
\left\|\begin{matrix} H_1(s) W_1(s) \\ H_2(s) W_2(s) \end{matrix}\right\|_\infty \le 1
|
||||
\end{equation}
|
||||
|
||||
Therefore, applying the $\mathcal{H}_\infty$ synthesis on the standard plant $P(s)$ eqref:eq:generalized_plant will generate two filters $H_2(s)$ and $H_1(s) \triangleq 1 - H_2(s)$ that are complementary eqref:eq:comp_filter_problem_form and such that there norms are bellow specified bounds [[eqref:eq:hinf_cond_h1]],eqref:eq:hinf_cond_h2.
|
||||
Therefore, applying the $\mathcal{H}_\infty$ synthesis to the standard plant $P(s)$ eqref:eq:generalized_plant will generate two filters $H_2(s)$ and $H_1(s) \triangleq 1 - H_2(s)$ that are complementary\nbsp{}eqref:eq:comp_filter_problem_form and such that there norms are bellow specified bounds [[eqref:eq:hinf_cond_h1]],\nbsp{}eqref:eq:hinf_cond_h2.
|
||||
|
||||
Note that there is not an equivalence between the $\mathcal{H}_\infty$ norm condition\nbsp{}eqref:eq:hinf_problem and the initial synthesis objectives\nbsp{}eqref:eq:hinf_cond_h1 and\nbsp{}eqref:eq:hinf_cond_h2, but only an implication.
|
||||
Hence, the optimization may be a little bit conservative with respect to the "set" of filters on which it is performed.
|
||||
There might be solutions were the objectives\nbsp{}eqref:eq:hinf_cond_h1 and\nbsp{}eqref:eq:hinf_cond_h2 are valid but where the $\mathcal{H}_\infty$ norm\nbsp{}eqref:eq:hinf_problem is larger than one.
|
||||
Note that there is only an implication between the $\mathcal{H}_\infty$ norm condition\nbsp{}eqref:eq:hinf_problem and the initial synthesis objectives\nbsp{}eqref:eq:hinf_cond_h1 and\nbsp{}eqref:eq:hinf_cond_h2 and not an equivalence.
|
||||
Hence, the optimization may be a little bit conservative with respect to the set of filters on which it is performed, see\nbsp{}[[cite:skogestad07_multiv_feedb_contr][Chap. 2.8.3]].
|
||||
In practice, this is however not an found to be an issue.
|
||||
|
||||
** Weighting Functions Design
|
||||
<<sec:hinf_weighting_func>>
|
||||
|
||||
Weighting functions are used during the synthesis to specify the maximum allowed norms of the complementary filters.
|
||||
Weighting functions are used during the synthesis to specify the maximum allowed complementary filters' norm.
|
||||
The proper design of these weighting functions is of primary importance for the success of the presented $\mathcal{H}_\infty$ synthesis of complementary filters.
|
||||
|
||||
First, only proper and stable transfer functions should be used.
|
||||
Second, the order of the weighting functions should stay reasonably small in order to reduce the computational costs associated with the solving of the optimization problem and for the physical implementation of the filters (the order of the synthesized filters being equal to the sum of the weighting functions orders).
|
||||
Third, one should not forget the fundamental limitations imposed by the complementary property eqref:eq:comp_filter.
|
||||
Second, the order of the weighting functions should stay reasonably small in order to reduce the computational costs associated with the solving of the optimization problem and for the physical implementation of the filters (the synthesized filters' order being equal to the sum of the weighting functions' order).
|
||||
Third, one should not forget the fundamental limitations imposed by the complementary property\nbsp{}eqref:eq:comp_filter.
|
||||
This implies for instance that $|H_1(j\omega)|$ and $|H_2(j\omega)|$ cannot be made small at the same frequency.
|
||||
|
||||
When designing complementary filters, it is usually desired to specify their slopes, their "blending" frequency and their maximum gains at low and high frequency.
|
||||
To help with the design of the weighting functions such that the above specification can be easily expressed, the formula eqref:eq:weight_formula is proposed.
|
||||
To easily express these specifications, formula\nbsp{}eqref:eq:weight_formula is proposed to help with the design of weighting functions.
|
||||
|
||||
#+name: eq:weight_formula
|
||||
\begin{equation}
|
||||
@@ -432,13 +439,13 @@ To help with the design of the weighting functions such that the above specifica
|
||||
}\right)^n
|
||||
\end{equation}
|
||||
|
||||
The parameters in formula eqref:eq:weight_formula are:
|
||||
- $G_0 = lim_{\omega \to 0} |W(j\omega)|$: the low frequency gain
|
||||
- $G_\infty = lim_{\omega \to \infty} |W(j\omega)|$: the high frequency gain
|
||||
The parameters in formula\nbsp{}eqref:eq:weight_formula are:
|
||||
- $G_0 = \lim_{\omega \to 0} |W(j\omega)|$: the low frequency gain
|
||||
- $G_\infty = \lim_{\omega \to \infty} |W(j\omega)|$: the high frequency gain
|
||||
- $G_c = |W(j\omega_c)|$: the gain at a specific frequency $\omega_c$ in $\si{rad/s}$.
|
||||
- $n$: the slope between high and low frequency. It also corresponds to the order of the weighting function.
|
||||
|
||||
The parameters $G_0$, $G_c$ and $G_\infty$ should either satisfy condition eqref:eq:cond_formula_1 or eqref:eq:cond_formula_2.
|
||||
The parameters $G_0$, $G_c$ and $G_\infty$ should either satisfy\nbsp{}eqref:eq:cond_formula_1 or\nbsp{}eqref:eq:cond_formula_2.
|
||||
#+name: eq:condition_params_formula
|
||||
\begin{subequations}
|
||||
\begin{align}
|
||||
@@ -447,22 +454,25 @@ The parameters $G_0$, $G_c$ and $G_\infty$ should either satisfy condition eqref
|
||||
\end{align}
|
||||
\end{subequations}
|
||||
|
||||
An example of the obtained magnitude of a weighting function generated using eqref:eq:weight_formula is shown in Fig.\nbsp{}ref:fig:weight_formula.
|
||||
The typical magnitude of a weighting function generated using\nbsp{}eqref:eq:weight_formula is shown in Fig.\nbsp{}ref:fig:weight_formula.
|
||||
|
||||
#+name: fig:weight_formula
|
||||
#+caption: Magnitude of a weighting function generated using the proposed formula eqref:eq:weight_formula, $G_0 = 1e^{-3}$, $G_\infty = 10$, $\omega_c = \SI{10}{Hz}$, $G_c = 2$, $n = 3$
|
||||
#+caption: Magnitude of a weighting function generated using formula\nbsp{}eqref:eq:weight_formula, $G_0 = 1e^{-3}$, $G_\infty = 10$, $\omega_c = \SI{10}{Hz}$, $G_c = 2$, $n = 3$.
|
||||
[[file:figs/weight_formula.pdf]]
|
||||
|
||||
** Validation of the proposed synthesis method
|
||||
<<sec:hinf_example>>
|
||||
|
||||
The proposed methodology for the design of complementary filters is now applied on a simple example where two complementary filters $H_1(s)$ and $H_2(s)$ have to be designed such that:
|
||||
- the blending frequency is around $\SI{10}{Hz}$
|
||||
- the slope of $|H_1(j\omega)|$ is $+2$ below $\SI{10}{Hz}$, its low frequency gain is $10^{-3}$
|
||||
- the slope of $|H_2(j\omega)|$ is $-3$ above $\SI{10}{Hz}$, its high frequency gain is $10^{-3}$
|
||||
The proposed methodology for the design of complementary filters is now applied on a simple example.
|
||||
Let's suppose two complementary filters $H_1(s)$ and $H_2(s)$ have to be designed such that:
|
||||
- the blending frequency is around $\SI{10}{Hz}$.
|
||||
- the slope of $|H_1(j\omega)|$ is $+2$ below $\SI{10}{Hz}$.
|
||||
Its low frequency gain is $10^{-3}$.
|
||||
- the slope of $|H_2(j\omega)|$ is $-3$ above $\SI{10}{Hz}$.
|
||||
Its high frequency gain is $10^{-3}$.
|
||||
|
||||
The first step is to translate the above requirements into the design of the weighting functions.
|
||||
The proposed formula eqref:eq:weight_formula is here used for such purpose.
|
||||
The first step is to translate the above requirements by properly designing the weighting functions.
|
||||
The proposed formula\nbsp{}eqref:eq:weight_formula is here used for such purpose.
|
||||
Parameters used are summarized in Table\nbsp{}ref:tab:weights_params.
|
||||
The inverse magnitudes of the designed weighting functions, which are representing the maximum allowed norms of the complementary filters, are shown by the dashed lines in Fig.\nbsp{}ref:fig:weights_W1_W2.
|
||||
|
||||
@@ -470,7 +480,7 @@ The inverse magnitudes of the designed weighting functions, which are representi
|
||||
\begin{figure}
|
||||
\begin{minipage}[b]{0.49\linewidth}
|
||||
\centering
|
||||
\begin{tabularx}{0.65\linewidth}{ccc}
|
||||
\begin{tabularx}{0.60\linewidth}{ccc}
|
||||
\toprule
|
||||
Parameters & \(W_1(s)\) & \(W_2(s)\)\\
|
||||
\midrule
|
||||
@@ -481,20 +491,20 @@ Parameters & \(W_1(s)\) & \(W_2(s)\)\\
|
||||
\(n\) & \(2\) & \(3\)\\
|
||||
\bottomrule
|
||||
\end{tabularx}
|
||||
\captionof{table}{\label{tab:weights_params}Parameters used for weighting functions \(W_1(s)\) and \(W_2(s)\) using \eqref{eq:weight_formula}}
|
||||
\captionof{table}{\label{tab:weights_params}Parameters used for \(W_1(s)\) and \(W_2(s)\) using~\eqref{eq:weight_formula}.}
|
||||
\end{minipage}
|
||||
\hfill
|
||||
\begin{minipage}[b]{0.49\linewidth}
|
||||
\centering
|
||||
\includegraphics[scale=1]{figs/weights_W1_W2.pdf}
|
||||
\caption{\label{fig:weights_W1_W2}Inverse magnitude of the weighting functions}
|
||||
\caption{\label{fig:weights_W1_W2}Inverse magnitude of the weighting functions.}
|
||||
\end{minipage}
|
||||
\end{figure}
|
||||
#+end_export
|
||||
|
||||
The $\mathcal{H}_\infty$ synthesis is then applied to the generalized plant of Fig.\nbsp{}ref:fig:h_infinity_robust_fusion_plant on efficiently solved in Matlab\nbsp{}cite:matlab20 using the Robust Control Toolbox.
|
||||
The standard $\mathcal{H}_\infty$ synthesis is then applied to the generalized plant of Fig.\nbsp{}ref:fig:h_infinity_robust_fusion_plant and efficiently solved using Matlab\nbsp{}cite:matlab20.
|
||||
The filter $H_2(s)$ that minimizes the $\mathcal{H}_\infty$ norm between $w$ and $[z_1,\ z_2]^T$ is obtained.
|
||||
The $\mathcal{H}_\infty$ norm is here found to be close to one eqref:eq:hinf_synthesis_result which indicates that the synthesis is successful: the complementary filters norms are below the maximum specified upper bounds.
|
||||
The $\mathcal{H}_\infty$ norm is here found to be close to one\nbsp{}eqref:eq:hinf_synthesis_result which indicates that the synthesis is successful: the complementary filters norms are below the maximum specified upper bounds.
|
||||
This is confirmed by the bode plots of the obtained complementary filters in Fig.\nbsp{}ref:fig:hinf_filters_results.
|
||||
|
||||
#+name: eq:hinf_synthesis_result
|
||||
@@ -502,8 +512,8 @@ This is confirmed by the bode plots of the obtained complementary filters in Fig
|
||||
\left\|\begin{matrix} \left(1 - H_2(s)\right) W_1(s) \\ H_2(s) W_2(s) \end{matrix}\right\|_\infty \approx 1
|
||||
\end{equation}
|
||||
|
||||
Their transfer functions in the Laplace domain are given in eqref:eq:hinf_synthesis_result_tf.
|
||||
As expected, the obtained filters are of order $5$, that is the sum of the weighting functions orders.
|
||||
The transfer functions in the Laplace domain of the complementary filters are given in\nbsp{}eqref:eq:hinf_synthesis_result_tf.
|
||||
As expected, the obtained filters are of order $5$, that is the sum of the weighting functions' order.
|
||||
|
||||
#+name: eq:hinf_synthesis_result_tf
|
||||
\begin{subequations}
|
||||
@@ -515,7 +525,7 @@ As expected, the obtained filters are of order $5$, that is the sum of the weigh
|
||||
|
||||
|
||||
#+name: fig:hinf_filters_results
|
||||
#+caption: Bode plot of the obtained complementary filters
|
||||
#+caption: Bode plot of the obtained complementary filters.
|
||||
[[file:figs/hinf_filters_results.pdf]]
|
||||
|
||||
This simple example illustrates the fact that the proposed methodology for complementary filters shaping is easy to use and effective.
|
||||
@@ -525,39 +535,39 @@ A more complex real life example is taken up in the next section.
|
||||
<<sec:application_ligo>>
|
||||
** Introduction :ignore:
|
||||
|
||||
Sensor fusion using complementary filters are widely used in active vibration isolation systems in gravitational wave detectors such at the LIGO\nbsp{}cite:matichard15_seism_isolat_advan_ligo,hua05_low_ligo, the VIRGO\nbsp{}cite:lucia18_low_frequen_optim_perfor_advan,heijningen18_low and the KAGRA [[cite:sekiguchi16_study_low_frequen_vibrat_isolat_system][Chap. 5]].
|
||||
Sensor fusion using complementary filters are widely used in the active vibration isolation systems at gravitational wave detectors, such as at the LIGO\nbsp{}cite:matichard15_seism_isolat_advan_ligo,hua05_low_ligo, the Virgo\nbsp{}cite:lucia18_low_frequen_optim_perfor_advan,heijningen18_low and the KAGRA [[cite:sekiguchi16_study_low_frequen_vibrat_isolat_system][Chap. 5]].
|
||||
|
||||
In the first isolation stage at the LIGO, two sets of complementary filters are used and included in a feedback loop\nbsp{}cite:hua04_low_ligo.
|
||||
In the first isolation stage at the LIGO, two sets of complementary filters are used to form a super sensor that is incorporated in a feedback loop\nbsp{}cite:hua04_low_ligo.
|
||||
A set of complementary filters ($L_2,H_2$) is first used to fuse a seismometer and a geophone.
|
||||
Then, another set of complementary filters ($L_1,H_1$) is used to merge the output of the first "inertial super sensor" with a position sensor.
|
||||
A simplified block diagram of the sensor fusion architecture is shown in Fig.\nbsp{}ref:fig:ligo_super_sensor_architecture.
|
||||
|
||||
#+name: fig:ligo_super_sensor_architecture
|
||||
#+caption: Simplified block diagram of the sensor blending strategy for the first stage at the LIGO\nbsp{}cite:hua04_low_ligo
|
||||
#+caption: Simplified block diagram of the sensor blending strategy for the first stage at the LIGO\nbsp{}cite:hua04_low_ligo.
|
||||
[[file:figs/ligo_super_sensor_architecture.pdf]]
|
||||
|
||||
The fusion of the position sensor at low frequency with the "inertial super sensor" at high frequency using the complementary filters ($L_1,H_1$) is done for several reasons, first of which is to give the super sensor a DC sensibility and therefore allow the feedback loop to have authority at zero frequency.
|
||||
The requirements on those filters are very tight and thus their design is complex and should be expressed as an optimization problem.
|
||||
The fusion of the position sensor at low frequency with the "inertial super sensor" at high frequency using the complementary filters ($L_1,H_1$) is done for several reasons, first of which is to give the super sensor a DC sensibility that allows the feedback loop to have authority at zero frequency.
|
||||
The requirements on those filters are stringent and thus their design is complex and should be expressed as an optimization problem.
|
||||
|
||||
The approach used in\nbsp{}cite:hua05_low_ligo is to use FIR complementary filters and to write the synthesis as a convex optimization problem.
|
||||
After synthesis, the obtained FIR filters were found to be compliant with the requirements.
|
||||
However they are of very high order so their implementation is quite complex.
|
||||
In this section, the effectiveness of the proposed complementary filter synthesis strategy is demonstrated on the same set of requirements.
|
||||
However they are of high order so their implementation is quite complex.
|
||||
In this section, the effectiveness of the proposed complementary filter synthesis strategy is demonstrated by using the same set of requirements.
|
||||
|
||||
** Complementary Filters Specifications
|
||||
<<sec:ligo_specifications>>
|
||||
|
||||
The specifications for the set of complementary filters ($L_1,H_1$) used at the LIGO are summarized below (for further details, refer to\nbsp{}cite:hua04_polyp_fir_compl_filter_contr_system):
|
||||
- From $0$ to $\SI{0.008}{Hz}$, the magnitude $|L_1(j\omega)|$ should be less or equal to $8 \times 10^{-4}$
|
||||
- Between $\SI{0.008}{Hz}$ to $\SI{0.04}{Hz}$, the filter $L_1(s)$ should attenuate the input signal proportional to frequency cubed
|
||||
- Between $\SI{0.04}{Hz}$ to $\SI{0.1}{Hz}$, the magnitude $|L_1(j\omega)|$ should be less than $3$
|
||||
- Above $\SI{0.1}{Hz}$, the magnitude $|H_1(j\omega)|$ should be less than $0.045$
|
||||
- Below $\SI{0.008}{Hz}$, the magnitude $|L_1(j\omega)|$ should be less than $8 \times 10^{-4}$.
|
||||
- From $\SI{0.008}{Hz}$ to $\SI{0.04}{Hz}$, the filter $L_1(s)$ should attenuate the input signal proportional to frequency cubed.
|
||||
- From $\SI{0.04}{Hz}$ to $\SI{0.1}{Hz}$, the magnitude $|L_1(j\omega)|$ should be less than $3$.
|
||||
- Above $\SI{0.1}{Hz}$, the magnitude $|H_1(j\omega)|$ should be less than $0.045$.
|
||||
|
||||
These specifications are therefore upper bounds on the complementary filters' magnitudes.
|
||||
These specifications are therefore upper bounds on the complementary filters' magnitude.
|
||||
They are physically represented in Fig.\nbsp{}ref:fig:fir_filter_ligo as well as the obtained magnitude of the FIR filters in\nbsp{}cite:hua05_low_ligo.
|
||||
|
||||
#+name: fig:fir_filter_ligo
|
||||
#+caption: Specifications and Bode plot of the obtained FIR complementary filters in\nbsp{}cite:hua05_low_ligo. The filters are here obtained using the SeDuMi Matlab toolbox\nbsp{}cite:sturm99_using_sedum
|
||||
#+caption: Specifications and Bode plot of the obtained FIR complementary filters in\nbsp{}cite:hua05_low_ligo. The filters are here obtained using the SeDuMi Matlab toolbox\nbsp{}cite:sturm99_using_sedum.
|
||||
[[file:figs/fir_filter_ligo.pdf]]
|
||||
|
||||
** Weighting Functions Design
|
||||
@@ -567,11 +577,11 @@ The weighting functions should be designed such that their inverse magnitude is
|
||||
However, the order of each weight should stay reasonably small in order to reduce the computational costs of the optimization problem as well as for the physical implementation of the filters.
|
||||
|
||||
A Type I Chebyshev filter of order $20$ is used for the weighting transfer function $W_L(s)$ corresponding to the low pass filter.
|
||||
For the one corresponding to the high pass filter $W_H(s)$, a $7^{\text{th}}$ order transfer function is designed.
|
||||
The magnitudes of the weighting functions are shown in Fig.\nbsp{}ref:fig:ligo_weights.
|
||||
For the one corresponding to the high pass filter $W_H(s)$, a $7^{\text{th}}$ order transfer function is manually designed.
|
||||
The inverse magnitudes of the weighting functions are shown in Fig.\nbsp{}ref:fig:ligo_weights.
|
||||
|
||||
#+name: fig:ligo_weights
|
||||
#+caption: Specifications and weighting functions inverse magnitudes
|
||||
#+caption: Specifications and weighting functions' inverse magnitude.
|
||||
[[file:figs/ligo_weights.pdf]]
|
||||
|
||||
** $\mathcal{H}_\infty$ Synthesis of the complementary filters
|
||||
@@ -580,54 +590,53 @@ The magnitudes of the weighting functions are shown in Fig.\nbsp{}ref:fig:ligo_w
|
||||
The proposed $\mathcal{H}_\infty$ synthesis is performed on the generalized plant shown in Fig.\nbsp{}ref:fig:h_infinity_robust_fusion_plant.
|
||||
After optimization, the $\mathcal{H}_\infty$ norm from $w$ to $[z_1,\ z_2]^T$ is found close to one indication successful synthesis.
|
||||
In Fig.\nbsp{}ref:fig:comp_fir_ligo_hinf, the bode plot of the obtained complementary filters are compared with the FIR filters of order 512 obtained in\nbsp{}cite:hua05_low_ligo.
|
||||
Even though the complementary filters using the $\mathcal{H}_\infty$ synthesis are of much lower order (order 27), they are found to be very close to the FIR filters.
|
||||
Even though the complementary filters using the $\mathcal{H}_\infty$ synthesis are of much lower order (order 27), they are found to be close to the FIR filters.
|
||||
This confirms the effectiveness of the proposed synthesis method even when the complementary filters are subject to complex requirements.
|
||||
|
||||
#+name: fig:comp_fir_ligo_hinf
|
||||
#+caption: Comparison of the FIR filters (dashed) designed in\nbsp{}cite:hua05_low_ligo with the filters obtained with $\mathcal{H}_\infty$ synthesis (solid)
|
||||
#+caption: Comparison of the FIR filters (dashed) designed in\nbsp{}cite:hua05_low_ligo with the filters obtained with $\mathcal{H}_\infty$ synthesis (solid).
|
||||
[[file:figs/comp_fir_ligo_hinf.pdf]]
|
||||
|
||||
* Discussion
|
||||
<<sec:discussion>>
|
||||
** Introduction :ignore:
|
||||
|
||||
** "Closed-Loop" complementary filters
|
||||
<<sec:closed_loop_complementary_filters>>
|
||||
|
||||
An alternative way to implement complementary filters is by using a fundamental property of the classical feedback architecture shown in Fig.\nbsp{}ref:fig:feedback_sensor_fusion.
|
||||
This is for instance presented in cite:mahony05_compl_filter_desig_special_orthog,plummer06_optim_compl_filter_their_applic_motion_measur,jensen13_basic_uas.
|
||||
This idea is discussed in\nbsp{}cite:mahony05_compl_filter_desig_special_orthog,plummer06_optim_compl_filter_their_applic_motion_measur,jensen13_basic_uas.
|
||||
|
||||
#+name: fig:feedback_sensor_fusion
|
||||
#+caption: "Closed-Loop" complementary filters
|
||||
#+caption: "Closed-Loop" complementary filters.
|
||||
[[file:figs/feedback_sensor_fusion.pdf]]
|
||||
|
||||
Consider the feedback architecture of Fig.\nbsp{}ref:fig:feedback_sensor_fusion, with two inputs $\hat{x}_1$ and $\hat{x}_2$, and one output $\hat{x}$.
|
||||
The output $\hat{x}$ is linked to the inputs by eqref:eq:closed_loop_complementary_filters.
|
||||
The output $\hat{x}$ is linked to the inputs by\nbsp{}eqref:eq:closed_loop_complementary_filters.
|
||||
|
||||
#+name: eq:closed_loop_complementary_filters
|
||||
\begin{equation}
|
||||
\hat{x} = \underbrace{\frac{1}{1 + L(s)}}_{S(s)} \hat{x}_1 + \underbrace{\frac{L(s)}{1 + L(s)}}_{T(s)} \hat{x}_2
|
||||
\end{equation}
|
||||
|
||||
As for any classical feedback architecture, we have that the sum of the sensitivity and complementary sensitivity transfer function is equal to one\nbsp{}eqref:eq:sensitivity_sum.
|
||||
As for any classical feedback architecture, we have that the sum of the sensitivity transfer function $S(s)$ and complementary sensitivity transfer function $T_(s)$ is equal to one\nbsp{}eqref:eq:sensitivity_sum.
|
||||
|
||||
#+name: eq:sensitivity_sum
|
||||
\begin{equation}
|
||||
T(s) + S(s) = 1
|
||||
S(s) + T(s) = 1
|
||||
\end{equation}
|
||||
|
||||
Therefore, provided that the closed-loop system is stable, the closed-loop system in Fig.\nbsp{}ref:fig:feedback_sensor_fusion is corresponding to two complementary filters.
|
||||
Therefore, provided that the the closed-loop system in Fig.\nbsp{}ref:fig:feedback_sensor_fusion is stable, it can be used as a set of two complementary filters.
|
||||
Two sensors can then be merged as shown in Fig.\nbsp{}ref:fig:feedback_sensor_fusion_arch.
|
||||
|
||||
#+name: fig:feedback_sensor_fusion_arch
|
||||
#+caption: Classical feedback architecture used for sensor fusion
|
||||
#+caption: Classical feedback architecture used for sensor fusion.
|
||||
[[file:figs/feedback_sensor_fusion_arch.pdf]]
|
||||
|
||||
One of the main advantage of implementing and designing complementary filters using the feedback architecture of Fig.\nbsp{}ref:fig:feedback_sensor_fusion is that all the tools of the linear control theory can be applied for the design of the filters.
|
||||
If one want to shape both $\frac{\hat{x}}{\hat{x}_1}(s) = S(s)$ and $\frac{\hat{x}}{\hat{x}_2}(s) = T(s)$, the $\mathcal{H}_\infty$ mixed-sensitivity synthesis can be easily applied.
|
||||
|
||||
To do so, weighting functions $W_1(s)$ and $W_2(s)$ are added to respectively shape $S(s)$ and $T(s)$ (Fig.\nbsp{}ref:fig:feedback_synthesis_architecture).
|
||||
Then the system is re-organized to form the generalized plant $P_L(s)$ shown in Fig.\nbsp{}ref:fig:feedback_synthesis_architecture_generalized_plant.
|
||||
The $\mathcal{H}_\infty$ mixed-sensitivity synthesis can finally be performed by applying the $\mathcal{H}_\infty$ synthesis to the generalized plant $P_L(s)$ which is described by eqref:eq:generalized_plant_mixed_sensitivity.
|
||||
Then the system is rearranged to form the generalized plant $P_L(s)$ shown in Fig.\nbsp{}ref:fig:feedback_synthesis_architecture_generalized_plant.
|
||||
The $\mathcal{H}_\infty$ mixed-sensitivity synthesis can finally be performed by applying the standard $\mathcal{H}_\infty$ synthesis to the generalized plant $P_L(s)$ which is described by\nbsp{}eqref:eq:generalized_plant_mixed_sensitivity.
|
||||
|
||||
#+name: eq:generalized_plant_mixed_sensitivity
|
||||
\begin{equation}
|
||||
@@ -637,63 +646,62 @@ The $\mathcal{H}_\infty$ mixed-sensitivity synthesis can finally be performed by
|
||||
\end{bmatrix}
|
||||
\end{equation}
|
||||
|
||||
The output of the synthesis is a filter $L(s)$ such that the "closed-loop" $\mathcal{H}_\infty$ norm from $[w_1,\ w_2]$ to $z$ is less than one eqref:eq:comp_filters_feedback_obj.
|
||||
The output of the synthesis is a filter $L(s)$ such that the "closed-loop" $\mathcal{H}_\infty$ norm from $[w_1,\ w_2]$ to $z$ of the system in Fig.\nbsp{}ref:fig:feedback_sensor_fusion is less than one\nbsp{}eqref:eq:comp_filters_feedback_obj.
|
||||
|
||||
#+name: eq:comp_filters_feedback_obj
|
||||
\begin{equation}
|
||||
\left\| \begin{matrix} \frac{z}{w_1} \\ \frac{z}{w_2} \end{matrix} \right\|_\infty = \left\| \begin{matrix} \frac{1}{1 + L(s)} W_1(s) \\ \frac{L(s)}{1 + L(s)} W_2(s) \end{matrix} \right\|_\infty \le 1
|
||||
\end{equation}
|
||||
|
||||
If the synthesis is successful, two complementary filters are obtained with their magnitudes bounded by the inverse magnitudes of the weighting functions.
|
||||
The sensor fusion can then be implemented as shown in Fig.\nbsp{}ref:fig:feedback_sensor_fusion_arch using the feedback architecture or more classically as shown in Fig.\nbsp{}ref:fig:sensor_fusion_overview by defining the two complementary filters as in\nbsp{}eqref:eq:comp_filters_feedback.
|
||||
If the synthesis is successful, the transfer functions from $\hat{x}_1$ to $\hat{x}$ and from $\hat{x}_2$ to $\hat{x}$ have their magnitude bounded by the inverse magnitude of the corresponding weighting functions.
|
||||
The sensor fusion can then be implemented using the feedback architecture in Fig.\nbsp{}ref:fig:feedback_sensor_fusion_arch or more classically as shown in Fig.\nbsp{}ref:fig:sensor_fusion_overview by defining the two complementary filters using\nbsp{}eqref:eq:comp_filters_feedback.
|
||||
The two architectures are equivalent regarding their inputs/outputs relationships.
|
||||
|
||||
#+name: eq:comp_filters_feedback
|
||||
\begin{equation}
|
||||
H_1(s) = \frac{1}{1 + L(s)}; \quad H_2(s) = \frac{L(s)}{1 + L(s)}
|
||||
\end{equation}
|
||||
|
||||
The two architecture are equivalent regarding their inputs/outputs relationships.
|
||||
|
||||
#+begin_export latex
|
||||
\begin{figure}[htbp]
|
||||
\begin{subfigure}[t]{0.6\linewidth}
|
||||
\centering
|
||||
\includegraphics[scale=1]{figs/feedback_synthesis_architecture.pdf}
|
||||
\caption{\label{fig:feedback_synthesis_architecture} Feedback architecture with included weights}
|
||||
\caption{\label{fig:feedback_synthesis_architecture} Feedback architecture with included weights.}
|
||||
\end{subfigure}
|
||||
\hfill
|
||||
\begin{subfigure}[t]{0.4\linewidth}
|
||||
\centering
|
||||
\includegraphics[scale=1]{figs/feedback_synthesis_architecture_generalized_plant.pdf}
|
||||
\caption{\label{fig:feedback_synthesis_architecture_generalized_plant} Generalized plant}
|
||||
\caption{\label{fig:feedback_synthesis_architecture_generalized_plant} Generalized plant.}
|
||||
\end{subfigure}
|
||||
\hfill
|
||||
\caption{\label{fig:h_inf_mixed_sensitivity_synthesis} $\mathcal{H}_\infty$ mixed-sensitivity synthesis}
|
||||
\caption{\label{fig:h_inf_mixed_sensitivity_synthesis} $\mathcal{H}_\infty$ mixed-sensitivity synthesis.}
|
||||
\centering
|
||||
\end{figure}
|
||||
#+end_export
|
||||
|
||||
As an example, two "closed-loop" complementary filters are designed using the $\mathcal{H}_\infty$ mixed-sensitivity synthesis.
|
||||
The weighting functions are designed using formula\nbsp{}eqref:eq:weight_formula with parameters shown in Table\nbsp{}ref:tab:weights_params.
|
||||
After synthesis, a filter $L(s)$ is obtained, its magnitude is shown in Fig.\nbsp{}ref:fig:hinf_filters_results_mixed_sensitivity by the dashed line.
|
||||
After synthesis, a filter $L(s)$ is obtained whose magnitude is shown in Fig.\nbsp{}ref:fig:hinf_filters_results_mixed_sensitivity by the black dashed line.
|
||||
The "closed-loop" complementary filters are compared with the inverse magnitude of the weighting functions in Fig.\nbsp{}ref:fig:hinf_filters_results_mixed_sensitivity confirming that the synthesis is successful.
|
||||
The obtained "closed-loop" complementary filters are indeed equal to the ones obtained in Section\nbsp{}ref:sec:hinf_example.
|
||||
|
||||
#+name: fig:hinf_filters_results_mixed_sensitivity
|
||||
#+caption: Bode plot of the obtained complementary filters after $\mathcal{H}_\infty$ mixed-sensitivity synthesis
|
||||
#+caption: Bode plot of the obtained complementary filters after $\mathcal{H}_\infty$ mixed-sensitivity synthesis.
|
||||
[[file:figs/hinf_filters_results_mixed_sensitivity.pdf]]
|
||||
|
||||
** Synthesis of more than two Complementary Filters
|
||||
** Synthesis of a set of three complementary filters
|
||||
<<sec:hinf_three_comp_filters>>
|
||||
|
||||
Some applications may require to merge more than two sensors\nbsp{}cite:stoten01_fusion_kinet_data_using_compos_filter,becker15_compl_filter_desig_three_frequen_bands.
|
||||
For instance at the LIGO\nbsp{}cite:matichard15_seism_isolat_advan_ligo, three sensors (an LVDT, a seismometer and a geophone) are merged to form a super sensor (Fig.\nbsp{}ref:fig:ligo_super_sensor_architecture).\par
|
||||
For instance at the LIGO, three sensors (an LVDT, a seismometer and a geophone) are merged to form a super sensor\nbsp{}cite:matichard15_seism_isolat_advan_ligo (Fig.\nbsp{}ref:fig:ligo_super_sensor_architecture).\par
|
||||
|
||||
When merging $n>2$ sensors using complementary filters, two architectures can be used as shown in Fig.\nbsp{}ref:fig:sensor_fusion_three.
|
||||
The fusion can either be done in a "sequential" way where $n-1$ sets of two complementary filters are used (Fig.\nbsp{}ref:fig:sensor_fusion_three_sequential), or in a "parallel" way where one set of $n$ complementary filters is used (Fig.\nbsp{}ref:fig:sensor_fusion_three_parallel).
|
||||
|
||||
In the first case, typical sensor fusion synthesis techniques can be used.
|
||||
However, when a parallel architecture is used, a new synthesis method for a set of more than two complementary filters is required as only simple analytical formulas have been proposed in the literature\nbsp{cite:stoten01_fusion_kinet_data_using_compos_filter,becker15_compl_filter_desig_three_frequen_bands}.
|
||||
However, when a parallel architecture is used, a new synthesis method for a set of more than two complementary filters is required as only simple analytical formulas have been proposed in the literature\nbsp{}cite:stoten01_fusion_kinet_data_using_compos_filter,becker15_compl_filter_desig_three_frequen_bands.
|
||||
A generalization of the proposed synthesis method of complementary filters is presented in this section. \par
|
||||
|
||||
#+begin_export latex
|
||||
@@ -701,15 +709,15 @@ A generalization of the proposed synthesis method of complementary filters is pr
|
||||
\begin{subfigure}[b]{0.59\linewidth}
|
||||
\centering
|
||||
\includegraphics[scale=1]{figs/sensor_fusion_three_sequential.pdf}
|
||||
\caption{\label{fig:sensor_fusion_three_sequential}Sequential fusion}
|
||||
\caption{\label{fig:sensor_fusion_three_sequential}Sequential fusion.}
|
||||
\end{subfigure}
|
||||
\hfill
|
||||
\begin{subfigure}[b]{0.39\linewidth}
|
||||
\centering
|
||||
\includegraphics[scale=1]{figs/sensor_fusion_three_parallel.pdf}
|
||||
\caption{\label{fig:sensor_fusion_three_parallel}Parallel fusion}
|
||||
\caption{\label{fig:sensor_fusion_three_parallel}Parallel fusion.}
|
||||
\end{subfigure}
|
||||
\caption{\label{fig:sensor_fusion_three}Possible sensor fusion architecture when more than two sensors are to be merged}
|
||||
\caption{\label{fig:sensor_fusion_three}Possible sensor fusion architecture when more than two sensors are to be merged.}
|
||||
\centering
|
||||
\end{figure}
|
||||
#+end_export
|
||||
@@ -718,62 +726,16 @@ The synthesis objective is to compute a set of $n$ stable transfer functions $[H
|
||||
#+name: eq:hinf_problem_gen
|
||||
\begin{subequations}
|
||||
\begin{align}
|
||||
& \sum_{i=0}^n H_i(s) = 1 \label{eq:hinf_cond_compl_gen} \\
|
||||
& \sum_{i=1}^n H_i(s) = 1 \label{eq:hinf_cond_compl_gen} \\
|
||||
& \left| H_i(j\omega) \right| < \frac{1}{\left| W_i(j\omega) \right|}, \quad \forall \omega,\ i = 1 \dots n \label{eq:hinf_cond_perf_gen}
|
||||
\end{align}
|
||||
\end{subequations}
|
||||
where $[W_1(s),\ W_2(s),\ \dots,\ W_n(s)]$ are weighting transfer functions that are chosen to specify the maximum wanted norms of the complementary filters during the synthesis.
|
||||
|
||||
Such synthesis objective is very close to the one described in Section\nbsp{}ref:sec:synthesis_objective, and indeed the proposed synthesis method is a generalization of the one presented in Section\nbsp{}ref:sec:hinf_synthesis. \par
|
||||
$[W_1(s),\ W_2(s),\ \dots,\ W_n(s)]$ are weighting transfer functions that are chosen to specify the maximum complementary filters' norm during the synthesis.
|
||||
|
||||
Before presenting the generalized synthesis method, the case with three sensors is presented.
|
||||
Consider the generalized plant $P_3(s)$ shown in Fig.\nbsp{}ref:fig:comp_filter_three_hinf_gen_plant which is also described by\nbsp{}eqref:eq:generalized_plant_three_filters.
|
||||
Such synthesis objective is closely related to the one described in Section\nbsp{}ref:sec:synthesis_objective, and indeed the proposed synthesis method is a generalization of the one presented in Section\nbsp{}ref:sec:hinf_synthesis. \par
|
||||
|
||||
#+name: eq:generalized_plant_three_filters
|
||||
\begin{equation}
|
||||
\begin{bmatrix} z_1 \\ z_2 \\ z_3 \\ v \end{bmatrix} = P_3(s) \begin{bmatrix} w \\ u_1 \\ u_2 \end{bmatrix}; \quad P_3(s) = \begin{bmatrix}W_1(s) & -W_1(s) & -W_1(s) \\ 0 & \phantom{+}W_2(s) & 0 \\ 0 & 0 & \phantom{+}W_3(s) \\ 1 & 0 & 0 \end{bmatrix}
|
||||
\end{equation}
|
||||
|
||||
#+begin_export latex
|
||||
\begin{figure}[htbp]
|
||||
\begin{subfigure}[t]{0.5\linewidth}
|
||||
\centering
|
||||
\includegraphics[scale=1]{figs/comp_filter_three_hinf_gen_plant.pdf}
|
||||
\caption{\label{fig:comp_filter_three_hinf_gen_plant} Generalized plant}
|
||||
\vfill
|
||||
\end{subfigure}
|
||||
\hfill
|
||||
\begin{subfigure}[t]{0.5\linewidth}
|
||||
\centering
|
||||
\includegraphics[scale=1]{figs/comp_filter_three_hinf_fb.pdf}
|
||||
\caption{\label{fig:comp_filter_three_hinf_fb} Generalized plant with the synthesized filter}
|
||||
\end{subfigure}
|
||||
\hfill
|
||||
\caption{\label{fig:comp_filter_three_hinf} Architecture for the $\mathcal{H}_\infty$ synthesis of three complementary filters}
|
||||
\centering
|
||||
\end{figure}
|
||||
#+end_export
|
||||
|
||||
Applying the $\mathcal{H}_\infty$ synthesis on the generalized plant $P_3(s)$ is equivalent as to find two stable filters $[H_2(s),\ H_3(s)]$ (shown in Fig.\nbsp{}ref:fig:comp_filter_three_hinf_fb) such that the $\mathcal{H}_\infty$ norm of the transfer function from $w$ to $[z_1,\ z_2, \ z_3]$ is less than one\nbsp{}eqref:eq:hinf_syn_obj_three.
|
||||
|
||||
#+name: eq:hinf_syn_obj_three
|
||||
\begin{equation}
|
||||
\left\| \begin{matrix} \left[1 - H_2(s) - H_3(s)\right] W_1(s) \\ H_2(s) W_2(s) \\ H_3(s) W_3(s) \end{matrix} \right\|_\infty \le 1
|
||||
\end{equation}
|
||||
|
||||
By defining $H_1(s) \triangleq 1 - H_2(s) - H_3(s)$, the proposed $\mathcal{H}_\infty$ synthesis solves the design problem eqref:eq:hinf_problem_gen with $n=3$. \par
|
||||
|
||||
An example is given to validate the method where three sensors are used in different frequency bands.
|
||||
For instance a displacement sensor from DC up to $\SI{1}{Hz}$, a geophone from $1$ to $\SI{10}{Hz}$ and an accelerometer above $\SI{10}{Hz}$.
|
||||
Three weighting functions are designed using formula\nbsp{}eqref:eq:weight_formula and their inverse magnitudes are shown in Fig.\nbsp{}ref:fig:three_complementary_filters_results (dashed curves).
|
||||
The $\mathcal{H}_\infty$ synthesis is performed on the generalized plant $P_3(s)$ and the bode plot of the obtained complementary filters are shown in Fig.\nbsp{}ref:fig:three_complementary_filters_results (solid lines). \par
|
||||
|
||||
#+name: fig:three_complementary_filters_results
|
||||
#+caption: Bode plot of the inverse weighting functions and of the three complementary filters obtained using the $\mathcal{H}_\infty$ synthesis
|
||||
[[file:figs/three_complementary_filters_results.pdf]]
|
||||
|
||||
Even though there might not be any practical application for a set of more than 3 complementary filters, it can still be designed using the same procedure.
|
||||
A set of $n$ complementary filters can be shaped using the generalized plant $P_n(s)$ described by\nbsp{}eqref:eq:generalized_plant_n_filters.
|
||||
A set of $n$ complementary filters can be shaped by applying the standard $\mathcal{H}_\infty$ synthesis to the generalized plant $P_n(s)$ described by\nbsp{}eqref:eq:generalized_plant_n_filters.
|
||||
|
||||
#+name: eq:generalized_plant_n_filters
|
||||
\begin{equation}
|
||||
@@ -788,17 +750,77 @@ A set of $n$ complementary filters can be shaped using the generalized plant $P_
|
||||
\end{bmatrix}
|
||||
\end{equation}
|
||||
|
||||
If the synthesis if successful, a set of $n-1$ filters $[H_2(s),\ H_3(s),\ \dots,\ H_n(s)]$ are obtained such that\nbsp{}eqref:eq:hinf_syn_obj_gen is verified.
|
||||
|
||||
#+name: eq:hinf_syn_obj_gen
|
||||
\begin{equation}
|
||||
\left\|\begin{matrix} \left(1 - \left[ H_2(s) + H_3(s) + \dots + H_n(s) \right]\right) W_1(s) \\ H_2(s) W_2(s) \\ \vdots \\ H_n(s) W_n(s) \end{matrix}\right\|_\infty \le 1
|
||||
\end{equation}
|
||||
|
||||
$H_1(s)$ is then defined using\nbsp{}eqref:eq:h1_comp_h2_hn which is ensuring the complementary property for the set of $n$ filters\nbsp{}eqref:eq:hinf_cond_compl_gen.
|
||||
Condition\nbsp{}eqref:eq:hinf_cond_perf_gen is satisfied thanks to\nbsp{}eqref:eq:hinf_syn_obj_gen.
|
||||
|
||||
#+name: eq:h1_comp_h2_hn
|
||||
\begin{equation}
|
||||
H_1(s) \triangleq 1 - \big[ H_2(s) + H_3(s) + \dots + H_n(s) \big]
|
||||
\end{equation}
|
||||
|
||||
An example is given to validate the proposed method for the synthesis of a set of three complementary filters.
|
||||
The sensors to be merged are a displacement sensor from DC up to $\SI{1}{Hz}$, a geophone from $1$ to $\SI{10}{Hz}$ and an accelerometer above $\SI{10}{Hz}$.
|
||||
Three weighting functions are designed using formula\nbsp{}eqref:eq:weight_formula and their inverse magnitude are shown in Fig.\nbsp{}ref:fig:three_complementary_filters_results (dashed curves).
|
||||
|
||||
Consider the generalized plant $P_3(s)$ shown in Fig.\nbsp{}ref:fig:comp_filter_three_hinf_gen_plant which is also described by\nbsp{}eqref:eq:generalized_plant_three_filters.
|
||||
|
||||
#+name: eq:generalized_plant_three_filters
|
||||
\begin{equation}
|
||||
\begin{bmatrix} z_1 \\ z_2 \\ z_3 \\ v \end{bmatrix} = P_3(s) \begin{bmatrix} w \\ u_1 \\ u_2 \end{bmatrix}; \quad P_3(s) = \begin{bmatrix}W_1(s) & -W_1(s) & -W_1(s) \\ 0 & \phantom{+}W_2(s) & 0 \\ 0 & 0 & \phantom{+}W_3(s) \\ 1 & 0 & 0 \end{bmatrix}
|
||||
\end{equation}
|
||||
|
||||
#+begin_export latex
|
||||
\begin{figure}[htbp]
|
||||
\begin{subfigure}[t]{0.5\linewidth}
|
||||
\centering
|
||||
\includegraphics[scale=1]{figs/comp_filter_three_hinf_gen_plant.pdf}
|
||||
\caption{\label{fig:comp_filter_three_hinf_gen_plant} Generalized plant.}
|
||||
\vfill
|
||||
\end{subfigure}
|
||||
\hfill
|
||||
\begin{subfigure}[t]{0.5\linewidth}
|
||||
\centering
|
||||
\includegraphics[scale=1]{figs/comp_filter_three_hinf_fb.pdf}
|
||||
\caption{\label{fig:comp_filter_three_hinf_fb} Generalized plant with the synthesized filter.}
|
||||
\end{subfigure}
|
||||
\hfill
|
||||
\caption{\label{fig:comp_filter_three_hinf} Architecture for the $\mathcal{H}_\infty$ synthesis of three complementary filters.}
|
||||
\centering
|
||||
\end{figure}
|
||||
#+end_export
|
||||
|
||||
The standard $\mathcal{H}_\infty$ synthesis is performed on the generalized plant $P_3(s)$.
|
||||
Two filters $H_2(s)$ and $H_3(s)$ are obtained such that the $\mathcal{H}_\infty$ norm of the closed-loop transfer from $w$ to $[z_1,\ z_2,\ z_3]$ of the system in Fig.\nbsp{}ref:fig:comp_filter_three_hinf_fb is less than one.
|
||||
Filter $H_1(s)$ is defined using\nbsp{}eqref:eq:h1_compl_h2_h3 thus ensuring the complementary property of the obtained set of filters.
|
||||
|
||||
#+name: eq:h1_compl_h2_h3
|
||||
\begin{equation}
|
||||
H_1(s) \triangleq 1 - \big[ H_2(s) + H_3(s) \big]
|
||||
\end{equation}
|
||||
|
||||
Figure\nbsp{}ref:fig:three_complementary_filters_results displays the three synthesized complementary filters (solid lines) which confirms that the synthesis is successful.\par
|
||||
|
||||
#+name: fig:three_complementary_filters_results
|
||||
#+caption: Bode plot of the inverse weighting functions and of the three complementary filters obtained using the $\mathcal{H}_\infty$ synthesis.
|
||||
[[file:figs/three_complementary_filters_results.pdf]]
|
||||
|
||||
* Conclusion
|
||||
<<sec:conclusion>>
|
||||
|
||||
Sensors measuring a physical quantities are always subject to limitations both in terms of bandwidth or accuracy.
|
||||
Complementary filters can be used to fuse multiple sensors with different characteristics in order to combine their benefits and yield a better estimate of the measured physical quantity.
|
||||
|
||||
The sensor fusion robustness and the obtained super sensor noise can be linked to the magnitude of the complementary filters.
|
||||
In this paper, a synthesis method that enables the shaping of the complementary filters norms has been proposed.
|
||||
Several example were used to emphasize the simplicity and the effectiveness of the proposed method.
|
||||
Links with "closed-loop" complementary filters where highlighted, and the proposed method was generalized for the design of a set of more than two complementary filters.
|
||||
A new method for designing complementary filters using the $\mathcal{H}_\infty$ synthesis has been proposed.
|
||||
It allows to shape the magnitude of the filters by the use of weighting functions during the synthesis.
|
||||
This is very valuable in practice as the characteristics of the super sensor are linked to the complementary filters' magnitude.
|
||||
Therefore typical sensor fusion objectives can be translated into requirements on the magnitudes of the filters.
|
||||
Several examples were used to emphasize the simplicity and the effectiveness of the proposed method.
|
||||
|
||||
However, the shaping of the complementary filters' magnitude does not allow to directly optimize the super sensor noise and dynamical characteristics.
|
||||
Future work will aim at developing a complementary filter synthesis method that minimizes the super sensor noise while ensuring the robustness of the fusion.
|
||||
|
||||
* Acknowledgment
|
||||
@@ -807,21 +829,20 @@ Future work will aim at developing a complementary filter synthesis method that
|
||||
:END:
|
||||
|
||||
This research benefited from a FRIA grant from the French Community of Belgium.
|
||||
This paper is based on a paper previously presented at the ICCMA conference\nbsp{}cite:dehaeze19_compl_filter_shapin_using_synth.
|
||||
This paper has been assigned the LIGO document number LIGO-P2100328.
|
||||
|
||||
* Data Availability
|
||||
:PROPERTIES:
|
||||
:UNNUMBERED: t
|
||||
:END:
|
||||
|
||||
Matlab cite:matlab20 was used for this study.
|
||||
Matlab\nbsp{}cite:matlab20 was used for this study.
|
||||
The source code is available under a MIT License and archived in Zenodo\nbsp{}cite:dehaeze21_new_method_desig_compl_filter_code.
|
||||
|
||||
* Bibliography :ignore:
|
||||
\bibliographystyle{elsarticle-num}
|
||||
\bibliography{ref}
|
||||
|
||||
|
||||
* Local Variables :noexport:
|
||||
# Local Variables:
|
||||
# org-latex-packages-alist: nil
|
||||
|
||||
Binary file not shown.
@@ -1,4 +1,4 @@
|
||||
% Created 2021-08-31 mar. 14:07
|
||||
% Created 2021-09-14 mar. 13:38
|
||||
% Intended LaTeX compiler: pdflatex
|
||||
\documentclass[preprint, sort&compress]{elsarticle}
|
||||
\usepackage[utf8]{inputenc}
|
||||
@@ -22,8 +22,8 @@
|
||||
\author[a3,a4]{Mohit Verma}
|
||||
\author[a2,a4]{Christophe Collette}
|
||||
\cortext[cor1]{Corresponding author. Email Address: thomas.dehaeze@esrf.fr}
|
||||
\address[a1]{European Synchrotron Radiation Facility, Grenoble, France}
|
||||
\address[a2]{University of Li\`{e}ge, Department of Aerospace and Mechanical Engineering, 4000 Li\`{e}ge, Belgium.}
|
||||
\address[a1]{European Synchrotron Radiation Facility, 38000 Grenoble, France}
|
||||
\address[a2]{University of Li\`{e}ge, PML, Department of Aerospace and Mechanical Engineering, 4000 Li\`{e}ge, Belgium.}
|
||||
\address[a3]{CSIR --- Structural Engineering Research Centre, Taramani, Chennai --- 600113, India.}
|
||||
\address[a4]{Universit\'{e} Libre de Bruxelles, Precision Mechatronics Laboratory, BEAMS Department, 1050 Brussels, Belgium.}
|
||||
\usepackage{tabularx}
|
||||
@@ -34,7 +34,7 @@
|
||||
\hypersetup{colorlinks=true}
|
||||
\setlength{\parskip}{1em}
|
||||
\date{}
|
||||
\title{A new method of designing complementary filters for sensor fusion using the \(\mathcal{H}_\infty\) synthesis}
|
||||
\title{Designing complementary filters for sensor fusion using \(\mathcal{H}_\infty\) synthesis}
|
||||
\begin{document}
|
||||
|
||||
|
||||
@@ -43,12 +43,13 @@
|
||||
\begin{frontmatter}
|
||||
\begin{abstract}
|
||||
In order to obtain a better estimate of a quantity being measured, several sensors having different characteristics can be merged with a technique called ``sensor fusion''.
|
||||
The obtained ``super sensor'' can combine the benefits of the individual sensors provided that the magnitude of the complementary filters used in the fusion are well shaped.
|
||||
Properly designing complementary filters is a difficult and time-consuming task.
|
||||
In this study, we address this issue and propose a new method for designing complementary filters.
|
||||
This method uses weighting functions to specify the wanted shape of the complementary filter that are then easily obtained using the standard \(\mathcal{H}_\infty\) synthesis on a specific generalized plant.
|
||||
The obtained ``super sensor'' combines the benefits of the individual sensors provided that the complementary filters used in the fusion are well designed.
|
||||
Indeed, properties of the super sensor are linked to the magnitude of the complementary filters.
|
||||
Properly shaping the magnitude of complementary filters is a difficult and time-consuming task.
|
||||
In this study, this issue is addressed and a new method for designing complementary filters is proposed.
|
||||
This method uses weighting functions to specify the wanted shape of the complementary filters that are then obtained using the standard \(\mathcal{H}_\infty\) synthesis.
|
||||
The proper choice of the weighting functions is discussed, and the effectiveness and simplicity of the design method is highlighted using several examples.
|
||||
Such synthesis method is further generalized to a set of more than two complementary filters.
|
||||
Such synthesis method is further extended for the shaping of a set of more than two complementary filters.
|
||||
\end{abstract}
|
||||
|
||||
\begin{keyword}
|
||||
@@ -57,76 +58,73 @@ Sensor fusion \sep{} Complementary filters \sep{} \(\mathcal{H}_\infty\) synthes
|
||||
\end{frontmatter}
|
||||
|
||||
\section{Introduction}
|
||||
\label{sec:org45b63f3}
|
||||
\label{sec:introduction}
|
||||
Measuring a physical quantity using sensors is always subject to several limitations.
|
||||
First, the accuracy of the measurement will be affected by several noise sources, such as the electrical noise of the conditioning electronics being used.
|
||||
First, the accuracy of the measurement is affected by several noise sources, such as electrical noise of the conditioning electronics being used.
|
||||
Second, the frequency range in which the measurement is relevant is bounded by the bandwidth of the sensor.
|
||||
One way to overcome these limitations is to combine several sensors using a technique called ``sensor fusion''~\cite{bendat57_optim_filter_indep_measur_two}.
|
||||
Fortunately, a wide variety of sensors exist, each with different characteristics.
|
||||
By carefully choosing the fused sensors, a so called ``super sensor'' is obtain that combines benefits of individual sensors and yields a better estimate of the measured physical quantity. \par
|
||||
Fortunately, a wide variety of sensors exists, each with different characteristics.
|
||||
By carefully choosing the fused sensors, a so called ``super sensor'' is obtained that can combines benefits of the individual sensors. \par
|
||||
In some situations, sensor fusion is used to increase the bandwidth of the measurement~\cite{shaw90_bandw_enhan_posit_measur_using_measur_accel,zimmermann92_high_bandw_orien_measur_contr,min15_compl_filter_desig_angle_estim}.
|
||||
For instance, in~\cite{shaw90_bandw_enhan_posit_measur_using_measur_accel}, the bandwidth of a position sensor is increased by fusing it with an accelerometer providing the high frequency motion information.
|
||||
For other applications, sensor fusion is used to obtain a estimate of the measured quantity with lower noise~\cite{hua05_low_ligo,hua04_polyp_fir_compl_filter_contr_system,plummer06_optim_compl_filter_their_applic_motion_measur,robert12_introd_random_signal_applied_kalman}.
|
||||
More recently, the fusion of sensors measuring different physical quantities has been proposed to obtained interesting properties for control~\cite{collette15_sensor_fusion_method_high_perfor,yong16_high_speed_vertic_posit_stage}.
|
||||
In~\cite{collette15_sensor_fusion_method_high_perfor}, an inertial sensor used for active vibration isolation is fused with a sensor collocated with the actuator to improve the stability margins of the feedback controller. \par
|
||||
For instance, in~\cite{shaw90_bandw_enhan_posit_measur_using_measur_accel} the bandwidth of a position sensor is increased by fusing it with an accelerometer providing the high frequency motion information.
|
||||
For other applications, sensor fusion is used to obtain an estimate of the measured quantity with lower noise~\cite{hua05_low_ligo,hua04_polyp_fir_compl_filter_contr_system,plummer06_optim_compl_filter_their_applic_motion_measur,robert12_introd_random_signal_applied_kalman}.
|
||||
More recently, the fusion of sensors measuring different physical quantities has been proposed to obtain interesting properties for control~\cite{collette15_sensor_fusion_method_high_perfor,yong16_high_speed_vertic_posit_stage}.
|
||||
In~\cite{collette15_sensor_fusion_method_high_perfor}, an inertial sensor used for active vibration isolation is fused with a sensor collocated with the actuator for improving the stability margins of the feedback controller. \par
|
||||
Practical applications of sensor fusion are numerous.
|
||||
It is widely used for the attitude estimation of several autonomous vehicles such as unmanned aerial vehicle~\cite{baerveldt97_low_cost_low_weigh_attit,corke04_inert_visual_sensin_system_small_auton_helic,jensen13_basic_uas} and underwater vehicles~\cite{pascoal99_navig_system_desig_using_time,batista10_optim_posit_veloc_navig_filter_auton_vehic}.
|
||||
Naturally, it is of great benefits for high performance positioning control as shown in~\cite{shaw90_bandw_enhan_posit_measur_using_measur_accel,zimmermann92_high_bandw_orien_measur_contr,min15_compl_filter_desig_angle_estim,yong16_high_speed_vertic_posit_stage}.
|
||||
Sensor fusion was also shown to be a key technology to improve the performances of active vibration isolation systems~\cite{tjepkema12_sensor_fusion_activ_vibrat_isolat_precis_equip}.
|
||||
This is particularly apparent for the isolation stages of gravitational wave observer~\cite{collette15_sensor_fusion_method_high_perfor,heijningen18_low} such as the ones used at the LIGO~\cite{hua05_low_ligo,hua04_polyp_fir_compl_filter_contr_system} and at the VIRGO~\cite{lucia18_low_frequen_optim_perfor_advan}. \par
|
||||
Sensor fusion was also shown to be a key technology to improve the performance of active vibration isolation systems~\cite{tjepkema12_sensor_fusion_activ_vibrat_isolat_precis_equip}.
|
||||
Emblematic examples are the isolation stages of gravitational wave detectors~\cite{collette15_sensor_fusion_method_high_perfor,heijningen18_low} such as the ones used at the LIGO~\cite{hua05_low_ligo,hua04_polyp_fir_compl_filter_contr_system} and at the Virgo~\cite{lucia18_low_frequen_optim_perfor_advan}. \par
|
||||
There are mainly two ways to perform sensor fusion: either using a set of complementary filters~\cite{anderson53_instr_approac_system_steer_comput} or using Kalman filtering~\cite{brown72_integ_navig_system_kalman_filter,odry18_kalman_filter_mobil_robot_attit_estim}.
|
||||
For sensor fusion applications, both methods are sharing many relationships~\cite{brown72_integ_navig_system_kalman_filter,higgins75_compar_compl_kalman_filter,robert12_introd_random_signal_applied_kalman,becker15_compl_filter_desig_three_frequen_bands}.
|
||||
However, for Kalman filtering, we are forced to make assumption about the probabilistic character of the sensor noises~\cite{robert12_introd_random_signal_applied_kalman} whereas it is not the case for complementary filters.
|
||||
Furthermore, the main advantages of complementary filters over Kalman filtering for sensor fusion are their very general applicability, their low computational cost~\cite{higgins75_compar_compl_kalman_filter}, and the fact that they are very intuitive as their effects can be easily interpreted in the frequency domain. \par
|
||||
However, for Kalman filtering, assumptions must be made about the probabilistic character of the sensor noises~\cite{robert12_introd_random_signal_applied_kalman} whereas it is not the case with complementary filters.
|
||||
Furthermore, the advantages of complementary filters over Kalman filtering for sensor fusion are their general applicability, their low computational cost~\cite{higgins75_compar_compl_kalman_filter}, and the fact that they are intuitive as their effects can be easily interpreted in the frequency domain. \par
|
||||
A set of filters is said to be complementary if the sum of their transfer functions is equal to one at all frequencies.
|
||||
For the earliest used of complementary filtering for sensor fusion, analog circuits were used to physically realize the filters~\cite{anderson53_instr_approac_system_steer_comput}.
|
||||
Analog complementary filters are still used today~\cite{yong16_high_speed_vertic_posit_stage,moore19_capac_instr_sensor_fusion_high_bandw_nanop}, but most of the time they are now implemented numerically as it allows for much more flexibility. \par
|
||||
In the early days of complementary filtering, analog circuits were employed to physically realize the filters~\cite{anderson53_instr_approac_system_steer_comput}.
|
||||
Analog complementary filters are still used today~\cite{yong16_high_speed_vertic_posit_stage,moore19_capac_instr_sensor_fusion_high_bandw_nanop}, but most of the time they are now implemented digitally as it allows for much more flexibility. \par
|
||||
|
||||
Several design methods have been developed over the years to optimize complementary filters.
|
||||
The easiest way to design complementary filters is to use analytical formulas.
|
||||
Depending on the application, the formulas used are of first order~\cite{corke04_inert_visual_sensin_system_small_auton_helic,yeh05_model_contr_hydraul_actuat_two,yong16_high_speed_vertic_posit_stage}, second order~\cite{baerveldt97_low_cost_low_weigh_attit,stoten01_fusion_kinet_data_using_compos_filter,jensen13_basic_uas} or even higher orders~\cite{shaw90_bandw_enhan_posit_measur_using_measur_accel,zimmermann92_high_bandw_orien_measur_contr,stoten01_fusion_kinet_data_using_compos_filter,collette15_sensor_fusion_method_high_perfor,matichard15_seism_isolat_advan_ligo}. \par
|
||||
|
||||
As the characteristics of the ``super sensor'' depends on the design of the complementary filters~\cite{dehaeze19_compl_filter_shapin_using_synth}, several optimization techniques were developed over the years.
|
||||
Some are based on the finding the optimal parameters of analytical formulas~\cite{jensen13_basic_uas,min15_compl_filter_desig_angle_estim,becker15_compl_filter_desig_three_frequen_bands}, while other are using convex optimization tools~\cite{hua04_polyp_fir_compl_filter_contr_system,hua05_low_ligo} such as linear matrix inequalities~\cite{pascoal99_navig_system_desig_using_time}.
|
||||
As the characteristics of the super sensor depends on the proper design of the complementary filters~\cite{dehaeze19_compl_filter_shapin_using_synth}, several optimization techniques have been developed.
|
||||
Some are based on the finding of optimal parameters of analytical formulas~\cite{jensen13_basic_uas,min15_compl_filter_desig_angle_estim,becker15_compl_filter_desig_three_frequen_bands}, while other are using convex optimization tools~\cite{hua04_polyp_fir_compl_filter_contr_system,hua05_low_ligo} such as linear matrix inequalities~\cite{pascoal99_navig_system_desig_using_time}.
|
||||
As shown in~\cite{plummer06_optim_compl_filter_their_applic_motion_measur}, the design of complementary filters can also be linked to the standard mixed-sensitivity control problem.
|
||||
Therefore, all the powerful tools developed for the classical control theory can also be used for the design of complementary filters.
|
||||
For instance, in~\cite{jensen13_basic_uas} the two gains of a Proportional Integral (PI) controller are optimized to minimize the noise of the super sensor. \par
|
||||
The common objective to all these complementary filters design methods is to obtain a super sensor that has desired characteristics, usually in terms of noise and dynamics.
|
||||
Moreover, as reported in~\cite{zimmermann92_high_bandw_orien_measur_contr,plummer06_optim_compl_filter_their_applic_motion_measur}, phase shifts and magnitude bumps of the ``super sensors'' dynamics can be observed if either the complementary filters are poorly designed or if the sensors are not well calibrated.
|
||||
Therefore, the robustness of the fusion is also of concerned when designing the complementary filters.
|
||||
Although many design methods of complementary filters have been proposed in the literature, no simple method that allows to specify the wanted super sensor characteristic while ensuring good fusion robustness has been proposed. \par
|
||||
The common objective of all these complementary filters design methods is to obtain a super sensor that has desired characteristics, usually in terms of noise and dynamics.
|
||||
Moreover, as reported in~\cite{zimmermann92_high_bandw_orien_measur_contr,plummer06_optim_compl_filter_their_applic_motion_measur}, phase shifts and magnitude bumps of the super sensors dynamics can be observed if either the complementary filters are poorly designed or if the sensors are not well calibrated.
|
||||
Hence, the robustness of the fusion is also of concern when designing the complementary filters.
|
||||
Although many design methods of complementary filters have been proposed in the literature, no simple method that allows to specify the desired super sensor characteristic while ensuring good fusion robustness has been proposed. \par
|
||||
Fortunately, both the robustness of the fusion and the super sensor characteristics can be linked to the magnitude of the complementary filters~\cite{dehaeze19_compl_filter_shapin_using_synth}.
|
||||
Based on that, this paper introduces a new way to design complementary filters using the \(\mathcal{H}_\infty\) synthesis which allows to shape the magnitude of the complementary filters in an easy and intuitive way. \par
|
||||
Section~\ref{sec:requirements} introduces the sensor fusion architecture and demonstrates how typical requirements can be linked to the complementary filters magnitudes.
|
||||
In section~\ref{sec:hinf_method}, the shaping of complementary filters is written as an \(\mathcal{H}_\infty\) optimization problem using weighting functions, and the simplicity of the proposed method is illustrated with an example.
|
||||
Based on that, this paper introduces a new way to design complementary filters using the \(\mathcal{H}_\infty\) synthesis which allows to shape the complementary filters' magnitude in an easy and intuitive way. \par
|
||||
Section~\ref{sec:requirements} introduces the sensor fusion architecture and demonstrates how typical requirements can be linked to the complementary filters' magnitude.
|
||||
In Section~\ref{sec:hinf_method}, the shaping of complementary filters is formulated as an \(\mathcal{H}_\infty\) optimization problem using weighting functions, and the simplicity of the proposed method is illustrated with an example.
|
||||
The synthesis method is further validated in Section~\ref{sec:application_ligo} by designing complex complementary filters.
|
||||
Section~\ref{sec:discussion} compares the proposed synthesis method with the classical mixed-sensitivity synthesis, and extends it to the shaping of more than two complementary filters.
|
||||
Section~\ref{sec:discussion} compares the proposed synthesis method with the classical mixed-sensitivity synthesis, and extends it for the shaping of more than two complementary filters.
|
||||
|
||||
\section{Sensor Fusion and Complementary Filters Requirements}
|
||||
\label{sec:orgeeb9584}
|
||||
\label{sec:requirements}
|
||||
Complementary filters provides a framework for fusing signals from different sensors.
|
||||
Complementary filtering provides a framework for fusing signals from different sensors.
|
||||
As the effectiveness of the fusion depends on the proper design of the complementary filters, they are expected to fulfill certain requirements.
|
||||
These requirements are discussed in this section.
|
||||
\subsection{Sensor Fusion Architecture}
|
||||
\label{sec:org556a17c}
|
||||
\label{sec:sensor_fusion}
|
||||
|
||||
A general sensor fusion architecture using complementary filters is shown in Fig.~\ref{fig:sensor_fusion_overview} where several sensors (here two) are measuring the same physical quantity \(x\).
|
||||
The two sensors output signals \(\hat{x}_1\) and \(\hat{x}_2\) are estimates of \(x\).
|
||||
Each of these estimates are then filtered out by complementary filters and combined to form a new estimate \(\hat{x}\).
|
||||
These estimates are then filtered out by complementary filters and combined to form a new estimate \(\hat{x}\).
|
||||
|
||||
The resulting sensor, termed as super sensor, can have larger bandwidth and better noise characteristics in comparison to the individual sensor.
|
||||
The resulting sensor, termed as super sensor, can have larger bandwidth and better noise characteristics in comparison to the individual sensors.
|
||||
This means that the super sensor provides an estimate \(\hat{x}\) of \(x\) which can be more accurate over a larger frequency band than the outputs of the individual sensors.
|
||||
|
||||
\begin{figure}[htbp]
|
||||
\centering
|
||||
\includegraphics[scale=1]{figs/sensor_fusion_overview.pdf}
|
||||
\caption{\label{fig:sensor_fusion_overview}Schematic of a sensor fusion architecture using complementary filters}
|
||||
\caption{\label{fig:sensor_fusion_overview}Schematic of a sensor fusion architecture using complementary filters.}
|
||||
\end{figure}
|
||||
|
||||
The complementary property of filters \(H_1(s)\) and \(H_2(s)\) implies that the sum of their transfer functions is equal to unity.
|
||||
The complementary property of filters \(H_1(s)\) and \(H_2(s)\) implies that the sum of their transfer functions is equal to one.
|
||||
That is, unity magnitude and zero phase at all frequencies.
|
||||
Therefore, a pair of complementary filter needs to satisfy the following condition:
|
||||
\begin{equation}
|
||||
@@ -137,7 +135,6 @@ Therefore, a pair of complementary filter needs to satisfy the following conditi
|
||||
It will soon become clear why the complementary property is important for the sensor fusion architecture.
|
||||
|
||||
\subsection{Sensor Models and Sensor Normalization}
|
||||
\label{sec:org5c5b6fb}
|
||||
\label{sec:sensor_models}
|
||||
|
||||
In order to study such sensor fusion architecture, a model for the sensors is required.
|
||||
@@ -148,21 +145,21 @@ Before filtering the sensor outputs \(\tilde{x}_i\) by the complementary filters
|
||||
This normalization consists of using an estimate \(\hat{G}_i(s)\) of the sensor dynamics \(G_i(s)\), and filtering the sensor output by the inverse of this estimate \(\hat{G}_i^{-1}(s)\) as shown in Fig.~\ref{fig:sensor_model_calibrated}.
|
||||
It is here supposed that the sensor inverse \(\hat{G}_i^{-1}(s)\) is proper and stable.
|
||||
This way, the units of the estimates \(\hat{x}_i\) are equal to the units of the physical quantity \(x\).
|
||||
The sensor dynamics estimate \(\hat{G}_1(s)\) can be a simple gain or a more complex transfer function.
|
||||
The sensor dynamics estimate \(\hat{G}_i(s)\) can be a simple gain or a more complex transfer function.
|
||||
|
||||
\begin{figure}[htbp]
|
||||
\begin{subfigure}[b]{0.49\linewidth}
|
||||
\centering
|
||||
\includegraphics[scale=1]{figs/sensor_model.pdf}
|
||||
\caption{\label{fig:sensor_model} Basic sensor model consisting of a noise input $n_i$ and a linear time invariant transfer function $G_i(s)$}
|
||||
\caption{\label{fig:sensor_model} Basic sensor model consisting of a noise input $n_i$ and a linear time invariant transfer function $G_i(s)$.}
|
||||
\end{subfigure}
|
||||
\hfill
|
||||
\begin{subfigure}[b]{0.49\linewidth}
|
||||
\centering
|
||||
\includegraphics[scale=1]{figs/sensor_model_calibrated.pdf}
|
||||
\caption{\label{fig:sensor_model_calibrated} Normalized sensors using the inverse of an estimate $\hat{G}_i(s)$ of the sensor dynamics}
|
||||
\caption{\label{fig:sensor_model_calibrated} Normalized sensors using the inverse of an estimate $\hat{G}_i(s)$ of the sensor dynamics.}
|
||||
\end{subfigure}
|
||||
\caption{\label{fig:sensor_models}Sensor models with and without normalization}
|
||||
\caption{\label{fig:sensor_models}Sensor models with and without normalization.}
|
||||
\centering
|
||||
\end{figure}
|
||||
|
||||
@@ -179,11 +176,10 @@ The super sensor output is therefore equal to:
|
||||
\begin{figure}[htbp]
|
||||
\centering
|
||||
\includegraphics[scale=1]{figs/fusion_super_sensor.pdf}
|
||||
\caption{\label{fig:fusion_super_sensor}Sensor fusion architecture with two normalized sensors}
|
||||
\caption{\label{fig:fusion_super_sensor}Sensor fusion architecture with two normalized sensors.}
|
||||
\end{figure}
|
||||
|
||||
\subsection{Noise Sensor Filtering}
|
||||
\label{sec:org4d93421}
|
||||
\label{sec:noise_filtering}
|
||||
|
||||
In this section, it is supposed that all the sensors are perfectly normalized, such that:
|
||||
@@ -194,23 +190,23 @@ In this section, it is supposed that all the sensors are perfectly normalized, s
|
||||
|
||||
The effect of a non-perfect normalization will be discussed in the next section.
|
||||
|
||||
Provided \eqref{eq:perfect_dynamics} is verified, the super sensor output \(\hat{x}\) is then:
|
||||
Provided~\eqref{eq:perfect_dynamics} is verified, the super sensor output \(\hat{x}\) is then equal to:
|
||||
\begin{equation}
|
||||
\label{eq:estimate_perfect_dyn}
|
||||
\hat{x} = x + H_1(s) n_1 + H_2(s) n_2
|
||||
\end{equation}
|
||||
|
||||
From \eqref{eq:estimate_perfect_dyn}, the complementary filters \(H_1(s)\) and \(H_2(s)\) are shown to only operate on the sensor's noises.
|
||||
From~\eqref{eq:estimate_perfect_dyn}, the complementary filters \(H_1(s)\) and \(H_2(s)\) are shown to only operate on the noise of the sensors.
|
||||
Thus, this sensor fusion architecture permits to filter the noise of both sensors without introducing any distortion in the physical quantity to be measured.
|
||||
This is why the two filters must be complementary.
|
||||
|
||||
The estimation error \(\delta x\), defined as the difference between the sensor output \(\hat{x}\) and the measured quantity \(x\), is computed for the super sensor \eqref{eq:estimate_error}.
|
||||
The estimation error \(\delta x\), defined as the difference between the sensor output \(\hat{x}\) and the measured quantity \(x\), is computed for the super sensor~\eqref{eq:estimate_error}.
|
||||
\begin{equation}
|
||||
\label{eq:estimate_error}
|
||||
\delta x \triangleq \hat{x} - x = H_1(s) n_1 + H_2(s) n_2
|
||||
\end{equation}
|
||||
|
||||
As shown in \eqref{eq:noise_filtering_psd}, the Power Spectral Density (PSD) of the estimation error \(\Phi_{\delta x}\) depends both on the norm of the two complementary filters and on the PSD of the noise sources \(\Phi_{n_1}\) and \(\Phi_{n_2}\).
|
||||
As shown in~\eqref{eq:noise_filtering_psd}, the Power Spectral Density (PSD) of the estimation error \(\Phi_{\delta x}\) depends both on the norm of the two complementary filters and on the PSD of the noise sources \(\Phi_{n_1}\) and \(\Phi_{n_2}\).
|
||||
\begin{equation}
|
||||
\label{eq:noise_filtering_psd}
|
||||
\Phi_{\delta x}(\omega) = \left|H_1(j\omega)\right|^2 \Phi_{n_1}(\omega) + \left|H_2(j\omega)\right|^2 \Phi_{n_2}(\omega)
|
||||
@@ -221,18 +217,17 @@ This is the simplest form of sensor fusion with complementary filters.
|
||||
|
||||
However, the two sensors have usually high noise levels over distinct frequency regions.
|
||||
In such case, to lower the noise of the super sensor, the norm \(|H_1(j\omega)|\) has to be small when \(\Phi_{n_1}(\omega)\) is larger than \(\Phi_{n_2}(\omega)\) and the norm \(|H_2(j\omega)|\) has to be small when \(\Phi_{n_2}(\omega)\) is larger than \(\Phi_{n_1}(\omega)\).
|
||||
Therefore, by properly shaping the norm of the complementary filters, it is possible to minimize the noise of the super sensor noise.
|
||||
Hence, by properly shaping the norm of the complementary filters, it is possible to reduce the noise of the super sensor.
|
||||
|
||||
\subsection{Sensor Fusion Robustness}
|
||||
\label{sec:org0ab9090}
|
||||
\label{sec:fusion_robustness}
|
||||
|
||||
In practical systems the sensor normalization is not perfect and condition \eqref{eq:perfect_dynamics} is not verified.
|
||||
In practical systems the sensor normalization is not perfect and condition~\eqref{eq:perfect_dynamics} is not verified.
|
||||
|
||||
In order to study such imperfection, a multiplicative input uncertainty is added to the sensor dynamics (Fig.~\ref{fig:sensor_model_uncertainty}).
|
||||
The nominal model is taken as the estimated model for the normalization \(\hat{G}_i(s)\), \(\Delta_i\) is any stable transfer function satisfying \(|\Delta_i(j\omega)| \le 1,\ \forall\omega\), and \(w_i(s)\) is a weighting transfer function representing the magnitude of the uncertainty.
|
||||
The nominal model is the estimated model used for the normalization \(\hat{G}_i(s)\), \(\Delta_i(s)\) is any stable transfer function satisfying \(|\Delta_i(j\omega)| \le 1,\ \forall\omega\), and \(w_i(s)\) is a weighting transfer function representing the magnitude of the uncertainty.
|
||||
|
||||
The weight \(w_i(s)\) is chosen such that the real sensor dynamics \(G(j\omega)\) is contained in the uncertain region represented by a circle in the complex plane, centered on \(1\) and with a radius equal to \(|w_i(j\omega)|\).
|
||||
The weight \(w_i(s)\) is chosen such that the real sensor dynamics \(G_i(j\omega)\) is contained in the uncertain region represented by a circle in the complex plane, centered on \(1\) and with a radius equal to \(|w_i(j\omega)|\).
|
||||
|
||||
As the nominal sensor dynamics is taken as the normalized filter, the normalized sensor can be further simplified as shown in Fig.~\ref{fig:sensor_model_uncertainty_simplified}.
|
||||
|
||||
@@ -240,34 +235,34 @@ As the nominal sensor dynamics is taken as the normalized filter, the normalized
|
||||
\begin{subfigure}[b]{0.59\linewidth}
|
||||
\centering
|
||||
\includegraphics[scale=1]{figs/sensor_model_uncertainty.pdf}
|
||||
\caption{\label{fig:sensor_model_uncertainty} Sensor with multiplicative input uncertainty}
|
||||
\caption{\label{fig:sensor_model_uncertainty} Sensor with multiplicative input uncertainty.}
|
||||
\end{subfigure}
|
||||
\hfill
|
||||
\begin{subfigure}[b]{0.39\linewidth}
|
||||
\centering
|
||||
\includegraphics[scale=1]{figs/sensor_model_uncertainty_simplified.pdf}
|
||||
\caption{\label{fig:sensor_model_uncertainty_simplified} Simplified sensor model}
|
||||
\caption{\label{fig:sensor_model_uncertainty_simplified} Simplified sensor model.}
|
||||
\end{subfigure}
|
||||
\hfill
|
||||
\caption{\label{fig:sensor_models_uncertainty}Sensor models with dynamical uncertainty}
|
||||
\caption{\label{fig:sensor_models_uncertainty}Sensor models with dynamical uncertainty.}
|
||||
\centering
|
||||
\end{figure}
|
||||
|
||||
The sensor fusion architecture with two sensor models including dynamical uncertainty is shown in Fig.~\ref{fig:sensor_fusion_dynamic_uncertainty}.
|
||||
The sensor fusion architecture with the sensor models including dynamical uncertainty is shown in Fig.~\ref{fig:sensor_fusion_dynamic_uncertainty}.
|
||||
|
||||
\begin{figure}[htbp]
|
||||
\centering
|
||||
\includegraphics[scale=1]{figs/sensor_fusion_dynamic_uncertainty.pdf}
|
||||
\caption{\label{fig:sensor_fusion_dynamic_uncertainty}Sensor fusion architecture with sensor dynamics uncertainty}
|
||||
\caption{\label{fig:sensor_fusion_dynamic_uncertainty}Sensor fusion architecture with sensor dynamics uncertainty.}
|
||||
\end{figure}
|
||||
|
||||
The super sensor dynamics \eqref{eq:super_sensor_dyn_uncertainty} is no longer equal to \(1\) and now depends on the sensor dynamical uncertainty weights \(w_i(s)\) as well as on the complementary filters \(H_i(s)\).
|
||||
The super sensor dynamics~\eqref{eq:super_sensor_dyn_uncertainty} is no longer equal to \(1\) and now depends on the sensor dynamical uncertainty weights \(w_i(s)\) as well as on the complementary filters \(H_i(s)\).
|
||||
\begin{equation}
|
||||
\label{eq:super_sensor_dyn_uncertainty}
|
||||
\frac{\hat{x}}{x} = 1 + w_1(s) H_1(s) \Delta_1(s) + w_2(s) H_2(s) \Delta_2(s)
|
||||
\end{equation}
|
||||
|
||||
The dynamical uncertainty of the super sensor can be graphically represented in the complex plane by a circle centered on \(1\) with a radius equal to \(|w_1(j\omega) H_1(j\omega)| + |w_2(j\omega) H_2(j\omega)|\) as shown in Fig.~\ref{fig:uncertainty_set_super_sensor}.
|
||||
The dynamical uncertainty of the super sensor can be graphically represented in the complex plane by a circle centered on \(1\) with a radius equal to \(|w_1(j\omega) H_1(j\omega)| + |w_2(j\omega) H_2(j\omega)|\) (Fig.~\ref{fig:uncertainty_set_super_sensor}).
|
||||
|
||||
\begin{figure}[htbp]
|
||||
\centering
|
||||
@@ -275,9 +270,8 @@ The dynamical uncertainty of the super sensor can be graphically represented in
|
||||
\caption{\label{fig:uncertainty_set_super_sensor}Uncertainty region of the super sensor dynamics in the complex plane (grey circle). The contribution of both sensors 1 and 2 to the total uncertainty are represented respectively by a blue circle and a red circle. The frequency dependency \(\omega\) is here omitted.}
|
||||
\end{figure}
|
||||
|
||||
The super sensor dynamical uncertainty (i.e. the robustness of the fusion) clearly depends on the complementary filters norms.
|
||||
For instance, the phase \(\Delta\phi(\omega)\) added by the super sensor dynamics at frequency \(\omega\) is bounded by \(\Delta\phi_{\text{max}}(\omega)\) which can be found by drawing a tangent from the origin to the uncertainty circle of the super sensor (Fig.~\ref{fig:uncertainty_set_super_sensor}).
|
||||
Therefore, the phase uncertainty of the super sensor dynamics depends on the Complementary filters norms \eqref{eq:max_phase_uncertainty}.
|
||||
The super sensor dynamical uncertainty, and hence the robustness of the fusion, clearly depends on the complementary filters' norm.
|
||||
For instance, the phase \(\Delta\phi(\omega)\) added by the super sensor dynamics at frequency \(\omega\) is bounded by \(\Delta\phi_{\text{max}}(\omega)\) which can be found by drawing a tangent from the origin to the uncertainty circle of the super sensor (Fig.~\ref{fig:uncertainty_set_super_sensor}) and that is mathematically described by~\eqref{eq:max_phase_uncertainty}.
|
||||
|
||||
\begin{equation}
|
||||
\label{eq:max_phase_uncertainty}
|
||||
@@ -288,19 +282,17 @@ As it is generally desired to limit the maximum phase added by the super sensor,
|
||||
Typically, the norm of the complementary filter \(|H_i(j\omega)|\) should be made small when \(|w_i(j\omega)|\) is large, i.e., at frequencies where the sensor dynamics is uncertain.
|
||||
|
||||
\section{Complementary Filters Shaping}
|
||||
\label{sec:orgdf903f0}
|
||||
\label{sec:hinf_method}
|
||||
As shown in Section~\ref{sec:requirements}, the noise and robustness of the super sensor are a function of the complementary filters norms.
|
||||
Therefore, a complementary filters synthesis method that allows to shape their norms would be of great use.
|
||||
As shown in Section~\ref{sec:requirements}, the noise and robustness of the super sensor are a function of the complementary filters' norm.
|
||||
Therefore, a synthesis method of complementary filters that allows to shape their norm would be of great use.
|
||||
In this section, such synthesis is proposed by writing the synthesis objective as a standard \(\mathcal{H}_\infty\) optimization problem.
|
||||
As weighting functions are used to represent the wanted complementary filters shapes during the synthesis, the proper design of weighting functions is discussed.
|
||||
As weighting functions are used to represent the wanted complementary filters' shape during the synthesis, their proper design is discussed.
|
||||
Finally, the synthesis method is validated on an simple example.
|
||||
\subsection{Synthesis Objective}
|
||||
\label{sec:org2467206}
|
||||
\label{sec:synthesis_objective}
|
||||
|
||||
The synthesis objective is to shape the norm of two filters \(H_1(s)\) and \(H_2(s)\) while ensuring their complementary property \eqref{eq:comp_filter}.
|
||||
This is equivalent as to finding proper and stable transfer functions \(H_1(s)\) and \(H_2(s)\) such that conditions \eqref{eq:hinf_cond_complementarity}, \eqref{eq:hinf_cond_h1} and \eqref{eq:hinf_cond_h2} are satisfied.
|
||||
The synthesis objective is to shape the norm of two filters \(H_1(s)\) and \(H_2(s)\) while ensuring their complementary property~\eqref{eq:comp_filter}.
|
||||
This is equivalent as to finding proper and stable transfer functions \(H_1(s)\) and \(H_2(s)\) such that conditions~\eqref{eq:hinf_cond_complementarity}, \eqref{eq:hinf_cond_h1} and~\eqref{eq:hinf_cond_h2} are satisfied.
|
||||
|
||||
\begin{subequations}
|
||||
\label{eq:comp_filter_problem_form}
|
||||
@@ -310,15 +302,14 @@ This is equivalent as to finding proper and stable transfer functions \(H_1(s)\)
|
||||
& |H_2(j\omega)| \le \frac{1}{|W_2(j\omega)|} \quad \forall\omega \label{eq:hinf_cond_h2}
|
||||
\end{align}
|
||||
\end{subequations}
|
||||
\(W_1(s)\) and \(W_2(s)\) are two weighting transfer functions that are carefully chosen to specify the maximum wanted norms of the complementary filters during the synthesis.
|
||||
\(W_1(s)\) and \(W_2(s)\) are two weighting transfer functions that are carefully chosen to specify the maximum wanted norm of the complementary filters during the synthesis.
|
||||
|
||||
\subsection{Shaping of Complementary Filters using \(\mathcal{H}_\infty\) synthesis}
|
||||
\label{sec:org172468e}
|
||||
\label{sec:hinf_synthesis}
|
||||
|
||||
In this section, it is shown that the synthesis objective can be easily expressed as a standard \(\mathcal{H}_\infty\) optimization problem and therefore solved using convenient tools readily available.
|
||||
|
||||
Consider the generalized plant \(P(s)\) shown in Fig.~\ref{fig:h_infinity_robust_fusion_plant} and mathematically described by \eqref{eq:generalized_plant}.
|
||||
Consider the generalized plant \(P(s)\) shown in Fig.~\ref{fig:h_infinity_robust_fusion_plant} and mathematically described by~\eqref{eq:generalized_plant}.
|
||||
|
||||
\begin{equation}
|
||||
\label{eq:generalized_plant}
|
||||
@@ -329,27 +320,27 @@ Consider the generalized plant \(P(s)\) shown in Fig.~\ref{fig:h_infinity_robust
|
||||
\begin{subfigure}[t]{0.5\linewidth}
|
||||
\centering
|
||||
\includegraphics[scale=1]{figs/h_infinity_robust_fusion_plant.pdf}
|
||||
\caption{\label{fig:h_infinity_robust_fusion_plant} Generalized plant}
|
||||
\caption{\label{fig:h_infinity_robust_fusion_plant} Generalized plant.}
|
||||
\vfill
|
||||
\end{subfigure}
|
||||
\hfill
|
||||
\begin{subfigure}[t]{0.5\linewidth}
|
||||
\centering
|
||||
\includegraphics[scale=1]{figs/h_infinity_robust_fusion_fb.pdf}
|
||||
\caption{\label{fig:h_infinity_robust_fusion_fb} Generalized plant with the synthesized filter}
|
||||
\caption{\label{fig:h_infinity_robust_fusion_fb} Generalized plant with the synthesized filter.}
|
||||
\end{subfigure}
|
||||
\hfill
|
||||
\caption{\label{fig:h_infinity_robust_fusion} Architecture for the $\mathcal{H}_\infty$ synthesis of complementary filters}
|
||||
\caption{\label{fig:h_infinity_robust_fusion} Architecture for the $\mathcal{H}_\infty$ synthesis of complementary filters.}
|
||||
\centering
|
||||
\end{figure}
|
||||
|
||||
Applying the standard \(\mathcal{H}_\infty\) synthesis on the generalized plant \(P(s)\) is then equivalent as finding a stable filter \(H_2(s)\) which based on \(v\), generates a signal \(u\) such that the \(\mathcal{H}_\infty\) norm of the system in Fig.~\ref{fig:h_infinity_robust_fusion_fb} from \(w\) to \([z_1, \ z_2]\) is less than one \eqref{eq:hinf_syn_obj}.
|
||||
Applying the standard \(\mathcal{H}_\infty\) synthesis to the generalized plant \(P(s)\) is then equivalent as finding a stable filter \(H_2(s)\) which based on \(v\), generates a signal \(u\) such that the \(\mathcal{H}_\infty\) norm of the system in Fig.~\ref{fig:h_infinity_robust_fusion_fb} from \(w\) to \([z_1, \ z_2]\) is less than one~\eqref{eq:hinf_syn_obj}.
|
||||
\begin{equation}
|
||||
\label{eq:hinf_syn_obj}
|
||||
\left\|\begin{matrix} \left(1 - H_2(s)\right) W_1(s) \\ H_2(s) W_2(s) \end{matrix}\right\|_\infty \le 1
|
||||
\end{equation}
|
||||
|
||||
By then defining \(H_1(s)\) to be the complementary of \(H_2(s)\) \eqref{eq:definition_H1}, the \(\mathcal{H}_\infty\) synthesis objective becomes equivalent to \eqref{eq:hinf_problem} which ensure that \eqref{eq:hinf_cond_h1} and \eqref{eq:hinf_cond_h2} are satisfied.
|
||||
By then defining \(H_1(s)\) to be the complementary of \(H_2(s)\) \eqref{eq:definition_H1}, the \(\mathcal{H}_\infty\) synthesis objective becomes equivalent to~\eqref{eq:hinf_problem} which ensures that~\eqref{eq:hinf_cond_h1} and~\eqref{eq:hinf_cond_h2} are satisfied.
|
||||
|
||||
\begin{equation}
|
||||
\label{eq:definition_H1}
|
||||
@@ -361,27 +352,25 @@ By then defining \(H_1(s)\) to be the complementary of \(H_2(s)\) \eqref{eq:defi
|
||||
\left\|\begin{matrix} H_1(s) W_1(s) \\ H_2(s) W_2(s) \end{matrix}\right\|_\infty \le 1
|
||||
\end{equation}
|
||||
|
||||
Therefore, applying the \(\mathcal{H}_\infty\) synthesis on the standard plant \(P(s)\) \eqref{eq:generalized_plant} will generate two filters \(H_2(s)\) and \(H_1(s) \triangleq 1 - H_2(s)\) that are complementary \eqref{eq:comp_filter_problem_form} and such that there norms are bellow specified bounds \eqref{eq:hinf_cond_h1},\eqref{eq:hinf_cond_h2}.
|
||||
Therefore, applying the \(\mathcal{H}_\infty\) synthesis to the standard plant \(P(s)\) \eqref{eq:generalized_plant} will generate two filters \(H_2(s)\) and \(H_1(s) \triangleq 1 - H_2(s)\) that are complementary~\eqref{eq:comp_filter_problem_form} and such that there norms are bellow specified bounds \eqref{eq:hinf_cond_h1},~\eqref{eq:hinf_cond_h2}.
|
||||
|
||||
Note that there is not an equivalence between the \(\mathcal{H}_\infty\) norm condition~\eqref{eq:hinf_problem} and the initial synthesis objectives~\eqref{eq:hinf_cond_h1} and~\eqref{eq:hinf_cond_h2}, but only an implication.
|
||||
Hence, the optimization may be a little bit conservative with respect to the ``set'' of filters on which it is performed.
|
||||
There might be solutions were the objectives~\eqref{eq:hinf_cond_h1} and~\eqref{eq:hinf_cond_h2} are valid but where the \(\mathcal{H}_\infty\) norm~\eqref{eq:hinf_problem} is larger than one.
|
||||
Note that there is only an implication between the \(\mathcal{H}_\infty\) norm condition~\eqref{eq:hinf_problem} and the initial synthesis objectives~\eqref{eq:hinf_cond_h1} and~\eqref{eq:hinf_cond_h2} and not an equivalence.
|
||||
Hence, the optimization may be a little bit conservative with respect to the set of filters on which it is performed, see~\cite[Chap. 2.8.3]{skogestad07_multiv_feedb_contr}.
|
||||
In practice, this is however not an found to be an issue.
|
||||
|
||||
\subsection{Weighting Functions Design}
|
||||
\label{sec:org3b7e958}
|
||||
\label{sec:hinf_weighting_func}
|
||||
|
||||
Weighting functions are used during the synthesis to specify the maximum allowed norms of the complementary filters.
|
||||
Weighting functions are used during the synthesis to specify the maximum allowed complementary filters' norm.
|
||||
The proper design of these weighting functions is of primary importance for the success of the presented \(\mathcal{H}_\infty\) synthesis of complementary filters.
|
||||
|
||||
First, only proper and stable transfer functions should be used.
|
||||
Second, the order of the weighting functions should stay reasonably small in order to reduce the computational costs associated with the solving of the optimization problem and for the physical implementation of the filters (the order of the synthesized filters being equal to the sum of the weighting functions orders).
|
||||
Third, one should not forget the fundamental limitations imposed by the complementary property \eqref{eq:comp_filter}.
|
||||
Second, the order of the weighting functions should stay reasonably small in order to reduce the computational costs associated with the solving of the optimization problem and for the physical implementation of the filters (the synthesized filters' order being equal to the sum of the weighting functions' order).
|
||||
Third, one should not forget the fundamental limitations imposed by the complementary property~\eqref{eq:comp_filter}.
|
||||
This implies for instance that \(|H_1(j\omega)|\) and \(|H_2(j\omega)|\) cannot be made small at the same frequency.
|
||||
|
||||
When designing complementary filters, it is usually desired to specify their slopes, their ``blending'' frequency and their maximum gains at low and high frequency.
|
||||
To help with the design of the weighting functions such that the above specification can be easily expressed, the formula \eqref{eq:weight_formula} is proposed.
|
||||
To easily express these specifications, formula~\eqref{eq:weight_formula} is proposed to help with the design of weighting functions.
|
||||
|
||||
\begin{equation}
|
||||
\label{eq:weight_formula}
|
||||
@@ -392,15 +381,15 @@ To help with the design of the weighting functions such that the above specifica
|
||||
}\right)^n
|
||||
\end{equation}
|
||||
|
||||
The parameters in formula \eqref{eq:weight_formula} are:
|
||||
The parameters in formula~\eqref{eq:weight_formula} are:
|
||||
\begin{itemize}
|
||||
\item \(G_0 = lim_{\omega \to 0} |W(j\omega)|\): the low frequency gain
|
||||
\item \(G_\infty = lim_{\omega \to \infty} |W(j\omega)|\): the high frequency gain
|
||||
\item \(G_0 = \lim_{\omega \to 0} |W(j\omega)|\): the low frequency gain
|
||||
\item \(G_\infty = \lim_{\omega \to \infty} |W(j\omega)|\): the high frequency gain
|
||||
\item \(G_c = |W(j\omega_c)|\): the gain at a specific frequency \(\omega_c\) in \(\si{rad/s}\).
|
||||
\item \(n\): the slope between high and low frequency. It also corresponds to the order of the weighting function.
|
||||
\end{itemize}
|
||||
|
||||
The parameters \(G_0\), \(G_c\) and \(G_\infty\) should either satisfy condition \eqref{eq:cond_formula_1} or \eqref{eq:cond_formula_2}.
|
||||
The parameters \(G_0\), \(G_c\) and \(G_\infty\) should either satisfy~\eqref{eq:cond_formula_1} or~\eqref{eq:cond_formula_2}.
|
||||
\begin{subequations}
|
||||
\label{eq:condition_params_formula}
|
||||
\begin{align}
|
||||
@@ -409,34 +398,36 @@ The parameters \(G_0\), \(G_c\) and \(G_\infty\) should either satisfy condition
|
||||
\end{align}
|
||||
\end{subequations}
|
||||
|
||||
An example of the obtained magnitude of a weighting function generated using \eqref{eq:weight_formula} is shown in Fig.~\ref{fig:weight_formula}.
|
||||
The typical magnitude of a weighting function generated using~\eqref{eq:weight_formula} is shown in Fig.~\ref{fig:weight_formula}.
|
||||
|
||||
\begin{figure}[htbp]
|
||||
\centering
|
||||
\includegraphics[scale=1]{figs/weight_formula.pdf}
|
||||
\caption{\label{fig:weight_formula}Magnitude of a weighting function generated using the proposed formula \eqref{eq:weight_formula}, \(G_0 = 1e^{-3}\), \(G_\infty = 10\), \(\omega_c = \SI{10}{Hz}\), \(G_c = 2\), \(n = 3\)}
|
||||
\caption{\label{fig:weight_formula}Magnitude of a weighting function generated using formula~\eqref{eq:weight_formula}, \(G_0 = 1e^{-3}\), \(G_\infty = 10\), \(\omega_c = \SI{10}{Hz}\), \(G_c = 2\), \(n = 3\).}
|
||||
\end{figure}
|
||||
|
||||
\subsection{Validation of the proposed synthesis method}
|
||||
\label{sec:org13db233}
|
||||
\label{sec:hinf_example}
|
||||
|
||||
The proposed methodology for the design of complementary filters is now applied on a simple example where two complementary filters \(H_1(s)\) and \(H_2(s)\) have to be designed such that:
|
||||
The proposed methodology for the design of complementary filters is now applied on a simple example.
|
||||
Let's suppose two complementary filters \(H_1(s)\) and \(H_2(s)\) have to be designed such that:
|
||||
\begin{itemize}
|
||||
\item the blending frequency is around \(\SI{10}{Hz}\)
|
||||
\item the slope of \(|H_1(j\omega)|\) is \(+2\) below \(\SI{10}{Hz}\), its low frequency gain is \(10^{-3}\)
|
||||
\item the slope of \(|H_2(j\omega)|\) is \(-3\) above \(\SI{10}{Hz}\), its high frequency gain is \(10^{-3}\)
|
||||
\item the blending frequency is around \(\SI{10}{Hz}\).
|
||||
\item the slope of \(|H_1(j\omega)|\) is \(+2\) below \(\SI{10}{Hz}\).
|
||||
Its low frequency gain is \(10^{-3}\).
|
||||
\item the slope of \(|H_2(j\omega)|\) is \(-3\) above \(\SI{10}{Hz}\).
|
||||
Its high frequency gain is \(10^{-3}\).
|
||||
\end{itemize}
|
||||
|
||||
The first step is to translate the above requirements into the design of the weighting functions.
|
||||
The proposed formula \eqref{eq:weight_formula} is here used for such purpose.
|
||||
The first step is to translate the above requirements by properly designing the weighting functions.
|
||||
The proposed formula~\eqref{eq:weight_formula} is here used for such purpose.
|
||||
Parameters used are summarized in Table~\ref{tab:weights_params}.
|
||||
The inverse magnitudes of the designed weighting functions, which are representing the maximum allowed norms of the complementary filters, are shown by the dashed lines in Fig.~\ref{fig:weights_W1_W2}.
|
||||
|
||||
\begin{figure}
|
||||
\begin{minipage}[b]{0.49\linewidth}
|
||||
\centering
|
||||
\begin{tabularx}{0.65\linewidth}{ccc}
|
||||
\begin{tabularx}{0.60\linewidth}{ccc}
|
||||
\toprule
|
||||
Parameters & \(W_1(s)\) & \(W_2(s)\)\\
|
||||
\midrule
|
||||
@@ -447,19 +438,19 @@ Parameters & \(W_1(s)\) & \(W_2(s)\)\\
|
||||
\(n\) & \(2\) & \(3\)\\
|
||||
\bottomrule
|
||||
\end{tabularx}
|
||||
\captionof{table}{\label{tab:weights_params}Parameters used for weighting functions \(W_1(s)\) and \(W_2(s)\) using \eqref{eq:weight_formula}}
|
||||
\captionof{table}{\label{tab:weights_params}Parameters used for \(W_1(s)\) and \(W_2(s)\) using~\eqref{eq:weight_formula}.}
|
||||
\end{minipage}
|
||||
\hfill
|
||||
\begin{minipage}[b]{0.49\linewidth}
|
||||
\centering
|
||||
\includegraphics[scale=1]{figs/weights_W1_W2.pdf}
|
||||
\caption{\label{fig:weights_W1_W2}Inverse magnitude of the weighting functions}
|
||||
\caption{\label{fig:weights_W1_W2}Inverse magnitude of the weighting functions.}
|
||||
\end{minipage}
|
||||
\end{figure}
|
||||
|
||||
The \(\mathcal{H}_\infty\) synthesis is then applied to the generalized plant of Fig.~\ref{fig:h_infinity_robust_fusion_plant} on efficiently solved in Matlab~\cite{matlab20} using the Robust Control Toolbox.
|
||||
The standard \(\mathcal{H}_\infty\) synthesis is then applied to the generalized plant of Fig.~\ref{fig:h_infinity_robust_fusion_plant} and efficiently solved using Matlab~\cite{matlab20}.
|
||||
The filter \(H_2(s)\) that minimizes the \(\mathcal{H}_\infty\) norm between \(w\) and \([z_1,\ z_2]^T\) is obtained.
|
||||
The \(\mathcal{H}_\infty\) norm is here found to be close to one \eqref{eq:hinf_synthesis_result} which indicates that the synthesis is successful: the complementary filters norms are below the maximum specified upper bounds.
|
||||
The \(\mathcal{H}_\infty\) norm is here found to be close to one~\eqref{eq:hinf_synthesis_result} which indicates that the synthesis is successful: the complementary filters norms are below the maximum specified upper bounds.
|
||||
This is confirmed by the bode plots of the obtained complementary filters in Fig.~\ref{fig:hinf_filters_results}.
|
||||
|
||||
\begin{equation}
|
||||
@@ -467,8 +458,8 @@ This is confirmed by the bode plots of the obtained complementary filters in Fig
|
||||
\left\|\begin{matrix} \left(1 - H_2(s)\right) W_1(s) \\ H_2(s) W_2(s) \end{matrix}\right\|_\infty \approx 1
|
||||
\end{equation}
|
||||
|
||||
Their transfer functions in the Laplace domain are given in \eqref{eq:hinf_synthesis_result_tf}.
|
||||
As expected, the obtained filters are of order \(5\), that is the sum of the weighting functions orders.
|
||||
The transfer functions in the Laplace domain of the complementary filters are given in~\eqref{eq:hinf_synthesis_result_tf}.
|
||||
As expected, the obtained filters are of order \(5\), that is the sum of the weighting functions' order.
|
||||
|
||||
\begin{subequations}
|
||||
\label{eq:hinf_synthesis_result_tf}
|
||||
@@ -482,18 +473,17 @@ As expected, the obtained filters are of order \(5\), that is the sum of the wei
|
||||
\begin{figure}[htbp]
|
||||
\centering
|
||||
\includegraphics[scale=1]{figs/hinf_filters_results.pdf}
|
||||
\caption{\label{fig:hinf_filters_results}Bode plot of the obtained complementary filters}
|
||||
\caption{\label{fig:hinf_filters_results}Bode plot of the obtained complementary filters.}
|
||||
\end{figure}
|
||||
|
||||
This simple example illustrates the fact that the proposed methodology for complementary filters shaping is easy to use and effective.
|
||||
A more complex real life example is taken up in the next section.
|
||||
|
||||
\section{Application: Design of Complementary Filters used in the Active Vibration Isolation System at the LIGO}
|
||||
\label{sec:orgb11268e}
|
||||
\label{sec:application_ligo}
|
||||
Sensor fusion using complementary filters are widely used in active vibration isolation systems in gravitational wave detectors such at the LIGO~\cite{matichard15_seism_isolat_advan_ligo,hua05_low_ligo}, the VIRGO~\cite{lucia18_low_frequen_optim_perfor_advan,heijningen18_low} and the KAGRA \cite[Chap. 5]{sekiguchi16_study_low_frequen_vibrat_isolat_system}.
|
||||
Sensor fusion using complementary filters are widely used in the active vibration isolation systems at gravitational wave detectors, such as at the LIGO~\cite{matichard15_seism_isolat_advan_ligo,hua05_low_ligo}, the Virgo~\cite{lucia18_low_frequen_optim_perfor_advan,heijningen18_low} and the KAGRA \cite[Chap. 5]{sekiguchi16_study_low_frequen_vibrat_isolat_system}.
|
||||
|
||||
In the first isolation stage at the LIGO, two sets of complementary filters are used and included in a feedback loop~\cite{hua04_low_ligo}.
|
||||
In the first isolation stage at the LIGO, two sets of complementary filters are used to form a super sensor that is incorporated in a feedback loop~\cite{hua04_low_ligo}.
|
||||
A set of complementary filters (\(L_2,H_2\)) is first used to fuse a seismometer and a geophone.
|
||||
Then, another set of complementary filters (\(L_1,H_1\)) is used to merge the output of the first ``inertial super sensor'' with a position sensor.
|
||||
A simplified block diagram of the sensor fusion architecture is shown in Fig.~\ref{fig:ligo_super_sensor_architecture}.
|
||||
@@ -501,115 +491,111 @@ A simplified block diagram of the sensor fusion architecture is shown in Fig.~\r
|
||||
\begin{figure}[htbp]
|
||||
\centering
|
||||
\includegraphics[scale=1]{figs/ligo_super_sensor_architecture.pdf}
|
||||
\caption{\label{fig:ligo_super_sensor_architecture}Simplified block diagram of the sensor blending strategy for the first stage at the LIGO~\cite{hua04_low_ligo}}
|
||||
\caption{\label{fig:ligo_super_sensor_architecture}Simplified block diagram of the sensor blending strategy for the first stage at the LIGO~\cite{hua04_low_ligo}.}
|
||||
\end{figure}
|
||||
|
||||
The fusion of the position sensor at low frequency with the ``inertial super sensor'' at high frequency using the complementary filters (\(L_1,H_1\)) is done for several reasons, first of which is to give the super sensor a DC sensibility and therefore allow the feedback loop to have authority at zero frequency.
|
||||
The requirements on those filters are very tight and thus their design is complex and should be expressed as an optimization problem.
|
||||
The fusion of the position sensor at low frequency with the ``inertial super sensor'' at high frequency using the complementary filters (\(L_1,H_1\)) is done for several reasons, first of which is to give the super sensor a DC sensibility that allows the feedback loop to have authority at zero frequency.
|
||||
The requirements on those filters are stringent and thus their design is complex and should be expressed as an optimization problem.
|
||||
|
||||
The approach used in~\cite{hua05_low_ligo} is to use FIR complementary filters and to write the synthesis as a convex optimization problem.
|
||||
After synthesis, the obtained FIR filters were found to be compliant with the requirements.
|
||||
However they are of very high order so their implementation is quite complex.
|
||||
In this section, the effectiveness of the proposed complementary filter synthesis strategy is demonstrated on the same set of requirements.
|
||||
However they are of high order so their implementation is quite complex.
|
||||
In this section, the effectiveness of the proposed complementary filter synthesis strategy is demonstrated by using the same set of requirements.
|
||||
\subsection{Complementary Filters Specifications}
|
||||
\label{sec:org100bb02}
|
||||
\label{sec:ligo_specifications}
|
||||
|
||||
The specifications for the set of complementary filters (\(L_1,H_1\)) used at the LIGO are summarized below (for further details, refer to~\cite{hua04_polyp_fir_compl_filter_contr_system}):
|
||||
\begin{itemize}
|
||||
\item From \(0\) to \(\SI{0.008}{Hz}\), the magnitude \(|L_1(j\omega)|\) should be less or equal to \(8 \times 10^{-4}\)
|
||||
\item Between \(\SI{0.008}{Hz}\) to \(\SI{0.04}{Hz}\), the filter \(L_1(s)\) should attenuate the input signal proportional to frequency cubed
|
||||
\item Between \(\SI{0.04}{Hz}\) to \(\SI{0.1}{Hz}\), the magnitude \(|L_1(j\omega)|\) should be less than \(3\)
|
||||
\item Above \(\SI{0.1}{Hz}\), the magnitude \(|H_1(j\omega)|\) should be less than \(0.045\)
|
||||
\item Below \(\SI{0.008}{Hz}\), the magnitude \(|L_1(j\omega)|\) should be less than \(8 \times 10^{-4}\).
|
||||
\item From \(\SI{0.008}{Hz}\) to \(\SI{0.04}{Hz}\), the filter \(L_1(s)\) should attenuate the input signal proportional to frequency cubed.
|
||||
\item From \(\SI{0.04}{Hz}\) to \(\SI{0.1}{Hz}\), the magnitude \(|L_1(j\omega)|\) should be less than \(3\).
|
||||
\item Above \(\SI{0.1}{Hz}\), the magnitude \(|H_1(j\omega)|\) should be less than \(0.045\).
|
||||
\end{itemize}
|
||||
|
||||
These specifications are therefore upper bounds on the complementary filters' magnitudes.
|
||||
These specifications are therefore upper bounds on the complementary filters' magnitude.
|
||||
They are physically represented in Fig.~\ref{fig:fir_filter_ligo} as well as the obtained magnitude of the FIR filters in~\cite{hua05_low_ligo}.
|
||||
|
||||
\begin{figure}[htbp]
|
||||
\centering
|
||||
\includegraphics[scale=1]{figs/fir_filter_ligo.pdf}
|
||||
\caption{\label{fig:fir_filter_ligo}Specifications and Bode plot of the obtained FIR complementary filters in~\cite{hua05_low_ligo}. The filters are here obtained using the SeDuMi Matlab toolbox~\cite{sturm99_using_sedum}}
|
||||
\caption{\label{fig:fir_filter_ligo}Specifications and Bode plot of the obtained FIR complementary filters in~\cite{hua05_low_ligo}. The filters are here obtained using the SeDuMi Matlab toolbox~\cite{sturm99_using_sedum}.}
|
||||
\end{figure}
|
||||
|
||||
\subsection{Weighting Functions Design}
|
||||
\label{sec:orgba233a6}
|
||||
\label{sec:ligo_weights}
|
||||
|
||||
The weighting functions should be designed such that their inverse magnitude is as close as possible to the specifications in order to not over-constrain the synthesis problem.
|
||||
However, the order of each weight should stay reasonably small in order to reduce the computational costs of the optimization problem as well as for the physical implementation of the filters.
|
||||
|
||||
A Type I Chebyshev filter of order \(20\) is used for the weighting transfer function \(W_L(s)\) corresponding to the low pass filter.
|
||||
For the one corresponding to the high pass filter \(W_H(s)\), a \(7^{\text{th}}\) order transfer function is designed.
|
||||
The magnitudes of the weighting functions are shown in Fig.~\ref{fig:ligo_weights}.
|
||||
For the one corresponding to the high pass filter \(W_H(s)\), a \(7^{\text{th}}\) order transfer function is manually designed.
|
||||
The inverse magnitudes of the weighting functions are shown in Fig.~\ref{fig:ligo_weights}.
|
||||
|
||||
\begin{figure}[htbp]
|
||||
\centering
|
||||
\includegraphics[scale=1]{figs/ligo_weights.pdf}
|
||||
\caption{\label{fig:ligo_weights}Specifications and weighting functions inverse magnitudes}
|
||||
\caption{\label{fig:ligo_weights}Specifications and weighting functions' inverse magnitude.}
|
||||
\end{figure}
|
||||
|
||||
\subsection{\(\mathcal{H}_\infty\) Synthesis of the complementary filters}
|
||||
\label{sec:org705f5c2}
|
||||
\label{sec:ligo_results}
|
||||
|
||||
The proposed \(\mathcal{H}_\infty\) synthesis is performed on the generalized plant shown in Fig.~\ref{fig:h_infinity_robust_fusion_plant}.
|
||||
After optimization, the \(\mathcal{H}_\infty\) norm from \(w\) to \([z_1,\ z_2]^T\) is found close to one indication successful synthesis.
|
||||
In Fig.~\ref{fig:comp_fir_ligo_hinf}, the bode plot of the obtained complementary filters are compared with the FIR filters of order 512 obtained in~\cite{hua05_low_ligo}.
|
||||
Even though the complementary filters using the \(\mathcal{H}_\infty\) synthesis are of much lower order (order 27), they are found to be very close to the FIR filters.
|
||||
Even though the complementary filters using the \(\mathcal{H}_\infty\) synthesis are of much lower order (order 27), they are found to be close to the FIR filters.
|
||||
This confirms the effectiveness of the proposed synthesis method even when the complementary filters are subject to complex requirements.
|
||||
|
||||
\begin{figure}[htbp]
|
||||
\centering
|
||||
\includegraphics[scale=1]{figs/comp_fir_ligo_hinf.pdf}
|
||||
\caption{\label{fig:comp_fir_ligo_hinf}Comparison of the FIR filters (dashed) designed in~\cite{hua05_low_ligo} with the filters obtained with \(\mathcal{H}_\infty\) synthesis (solid)}
|
||||
\caption{\label{fig:comp_fir_ligo_hinf}Comparison of the FIR filters (dashed) designed in~\cite{hua05_low_ligo} with the filters obtained with \(\mathcal{H}_\infty\) synthesis (solid).}
|
||||
\end{figure}
|
||||
|
||||
\section{Discussion}
|
||||
\label{sec:org2d41692}
|
||||
\label{sec:discussion}
|
||||
\subsection{``Closed-Loop'' complementary filters}
|
||||
\label{sec:org2106ef4}
|
||||
\label{sec:closed_loop_complementary_filters}
|
||||
|
||||
An alternative way to implement complementary filters is by using a fundamental property of the classical feedback architecture shown in Fig.~\ref{fig:feedback_sensor_fusion}.
|
||||
This is for instance presented in \cite{mahony05_compl_filter_desig_special_orthog,plummer06_optim_compl_filter_their_applic_motion_measur,jensen13_basic_uas}.
|
||||
This idea is discussed in~\cite{mahony05_compl_filter_desig_special_orthog,plummer06_optim_compl_filter_their_applic_motion_measur,jensen13_basic_uas}.
|
||||
|
||||
\begin{figure}[htbp]
|
||||
\centering
|
||||
\includegraphics[scale=1]{figs/feedback_sensor_fusion.pdf}
|
||||
\caption{\label{fig:feedback_sensor_fusion}``Closed-Loop'' complementary filters}
|
||||
\caption{\label{fig:feedback_sensor_fusion}``Closed-Loop'' complementary filters.}
|
||||
\end{figure}
|
||||
|
||||
Consider the feedback architecture of Fig.~\ref{fig:feedback_sensor_fusion}, with two inputs \(\hat{x}_1\) and \(\hat{x}_2\), and one output \(\hat{x}\).
|
||||
The output \(\hat{x}\) is linked to the inputs by \eqref{eq:closed_loop_complementary_filters}.
|
||||
The output \(\hat{x}\) is linked to the inputs by~\eqref{eq:closed_loop_complementary_filters}.
|
||||
|
||||
\begin{equation}
|
||||
\label{eq:closed_loop_complementary_filters}
|
||||
\hat{x} = \underbrace{\frac{1}{1 + L(s)}}_{S(s)} \hat{x}_1 + \underbrace{\frac{L(s)}{1 + L(s)}}_{T(s)} \hat{x}_2
|
||||
\end{equation}
|
||||
|
||||
As for any classical feedback architecture, we have that the sum of the sensitivity and complementary sensitivity transfer function is equal to one~\eqref{eq:sensitivity_sum}.
|
||||
As for any classical feedback architecture, we have that the sum of the sensitivity transfer function \(S(s)\) and complementary sensitivity transfer function \(T_(s)\) is equal to one~\eqref{eq:sensitivity_sum}.
|
||||
|
||||
\begin{equation}
|
||||
\label{eq:sensitivity_sum}
|
||||
T(s) + S(s) = 1
|
||||
S(s) + T(s) = 1
|
||||
\end{equation}
|
||||
|
||||
Therefore, provided that the closed-loop system is stable, the closed-loop system in Fig.~\ref{fig:feedback_sensor_fusion} is corresponding to two complementary filters.
|
||||
Therefore, provided that the the closed-loop system in Fig.~\ref{fig:feedback_sensor_fusion} is stable, it can be used as a set of two complementary filters.
|
||||
Two sensors can then be merged as shown in Fig.~\ref{fig:feedback_sensor_fusion_arch}.
|
||||
|
||||
\begin{figure}[htbp]
|
||||
\centering
|
||||
\includegraphics[scale=1]{figs/feedback_sensor_fusion_arch.pdf}
|
||||
\caption{\label{fig:feedback_sensor_fusion_arch}Classical feedback architecture used for sensor fusion}
|
||||
\caption{\label{fig:feedback_sensor_fusion_arch}Classical feedback architecture used for sensor fusion.}
|
||||
\end{figure}
|
||||
|
||||
One of the main advantage of implementing and designing complementary filters using the feedback architecture of Fig.~\ref{fig:feedback_sensor_fusion} is that all the tools of the linear control theory can be applied for the design of the filters.
|
||||
If one want to shape both \(\frac{\hat{x}}{\hat{x}_1}(s) = S(s)\) and \(\frac{\hat{x}}{\hat{x}_2}(s) = T(s)\), the \(\mathcal{H}_\infty\) mixed-sensitivity synthesis can be easily applied.
|
||||
|
||||
To do so, weighting functions \(W_1(s)\) and \(W_2(s)\) are added to respectively shape \(S(s)\) and \(T(s)\) (Fig.~\ref{fig:feedback_synthesis_architecture}).
|
||||
Then the system is re-organized to form the generalized plant \(P_L(s)\) shown in Fig.~\ref{fig:feedback_synthesis_architecture_generalized_plant}.
|
||||
The \(\mathcal{H}_\infty\) mixed-sensitivity synthesis can finally be performed by applying the \(\mathcal{H}_\infty\) synthesis to the generalized plant \(P_L(s)\) which is described by \eqref{eq:generalized_plant_mixed_sensitivity}.
|
||||
Then the system is rearranged to form the generalized plant \(P_L(s)\) shown in Fig.~\ref{fig:feedback_synthesis_architecture_generalized_plant}.
|
||||
The \(\mathcal{H}_\infty\) mixed-sensitivity synthesis can finally be performed by applying the standard \(\mathcal{H}_\infty\) synthesis to the generalized plant \(P_L(s)\) which is described by~\eqref{eq:generalized_plant_mixed_sensitivity}.
|
||||
|
||||
\begin{equation}
|
||||
\label{eq:generalized_plant_mixed_sensitivity}
|
||||
@@ -619,79 +605,77 @@ The \(\mathcal{H}_\infty\) mixed-sensitivity synthesis can finally be performed
|
||||
\end{bmatrix}
|
||||
\end{equation}
|
||||
|
||||
The output of the synthesis is a filter \(L(s)\) such that the ``closed-loop'' \(\mathcal{H}_\infty\) norm from \([w_1,\ w_2]\) to \(z\) is less than one \eqref{eq:comp_filters_feedback_obj}.
|
||||
The output of the synthesis is a filter \(L(s)\) such that the ``closed-loop'' \(\mathcal{H}_\infty\) norm from \([w_1,\ w_2]\) to \(z\) of the system in Fig.~\ref{fig:feedback_sensor_fusion} is less than one~\eqref{eq:comp_filters_feedback_obj}.
|
||||
|
||||
\begin{equation}
|
||||
\label{eq:comp_filters_feedback_obj}
|
||||
\left\| \begin{matrix} \frac{z}{w_1} \\ \frac{z}{w_2} \end{matrix} \right\|_\infty = \left\| \begin{matrix} \frac{1}{1 + L(s)} W_1(s) \\ \frac{L(s)}{1 + L(s)} W_2(s) \end{matrix} \right\|_\infty \le 1
|
||||
\end{equation}
|
||||
|
||||
If the synthesis is successful, two complementary filters are obtained with their magnitudes bounded by the inverse magnitudes of the weighting functions.
|
||||
The sensor fusion can then be implemented as shown in Fig.~\ref{fig:feedback_sensor_fusion_arch} using the feedback architecture or more classically as shown in Fig.~\ref{fig:sensor_fusion_overview} by defining the two complementary filters as in~\eqref{eq:comp_filters_feedback}.
|
||||
If the synthesis is successful, the transfer functions from \(\hat{x}_1\) to \(\hat{x}\) and from \(\hat{x}_2\) to \(\hat{x}\) have their magnitude bounded by the inverse magnitude of the corresponding weighting functions.
|
||||
The sensor fusion can then be implemented using the feedback architecture in Fig.~\ref{fig:feedback_sensor_fusion_arch} or more classically as shown in Fig.~\ref{fig:sensor_fusion_overview} by defining the two complementary filters using~\eqref{eq:comp_filters_feedback}.
|
||||
The two architectures are equivalent regarding their inputs/outputs relationships.
|
||||
|
||||
\begin{equation}
|
||||
\label{eq:comp_filters_feedback}
|
||||
H_1(s) = \frac{1}{1 + L(s)}; \quad H_2(s) = \frac{L(s)}{1 + L(s)}
|
||||
\end{equation}
|
||||
|
||||
The two architecture are equivalent regarding their inputs/outputs relationships.
|
||||
|
||||
\begin{figure}[htbp]
|
||||
\begin{subfigure}[t]{0.6\linewidth}
|
||||
\centering
|
||||
\includegraphics[scale=1]{figs/feedback_synthesis_architecture.pdf}
|
||||
\caption{\label{fig:feedback_synthesis_architecture} Feedback architecture with included weights}
|
||||
\caption{\label{fig:feedback_synthesis_architecture} Feedback architecture with included weights.}
|
||||
\end{subfigure}
|
||||
\hfill
|
||||
\begin{subfigure}[t]{0.4\linewidth}
|
||||
\centering
|
||||
\includegraphics[scale=1]{figs/feedback_synthesis_architecture_generalized_plant.pdf}
|
||||
\caption{\label{fig:feedback_synthesis_architecture_generalized_plant} Generalized plant}
|
||||
\caption{\label{fig:feedback_synthesis_architecture_generalized_plant} Generalized plant.}
|
||||
\end{subfigure}
|
||||
\hfill
|
||||
\caption{\label{fig:h_inf_mixed_sensitivity_synthesis} $\mathcal{H}_\infty$ mixed-sensitivity synthesis}
|
||||
\caption{\label{fig:h_inf_mixed_sensitivity_synthesis} $\mathcal{H}_\infty$ mixed-sensitivity synthesis.}
|
||||
\centering
|
||||
\end{figure}
|
||||
|
||||
As an example, two ``closed-loop'' complementary filters are designed using the \(\mathcal{H}_\infty\) mixed-sensitivity synthesis.
|
||||
The weighting functions are designed using formula~\eqref{eq:weight_formula} with parameters shown in Table~\ref{tab:weights_params}.
|
||||
After synthesis, a filter \(L(s)\) is obtained, its magnitude is shown in Fig.~\ref{fig:hinf_filters_results_mixed_sensitivity} by the dashed line.
|
||||
After synthesis, a filter \(L(s)\) is obtained whose magnitude is shown in Fig.~\ref{fig:hinf_filters_results_mixed_sensitivity} by the black dashed line.
|
||||
The ``closed-loop'' complementary filters are compared with the inverse magnitude of the weighting functions in Fig.~\ref{fig:hinf_filters_results_mixed_sensitivity} confirming that the synthesis is successful.
|
||||
The obtained ``closed-loop'' complementary filters are indeed equal to the ones obtained in Section~\ref{sec:hinf_example}.
|
||||
|
||||
\begin{figure}[htbp]
|
||||
\centering
|
||||
\includegraphics[scale=1]{figs/hinf_filters_results_mixed_sensitivity.pdf}
|
||||
\caption{\label{fig:hinf_filters_results_mixed_sensitivity}Bode plot of the obtained complementary filters after \(\mathcal{H}_\infty\) mixed-sensitivity synthesis}
|
||||
\caption{\label{fig:hinf_filters_results_mixed_sensitivity}Bode plot of the obtained complementary filters after \(\mathcal{H}_\infty\) mixed-sensitivity synthesis.}
|
||||
\end{figure}
|
||||
|
||||
\subsection{Synthesis of more than two Complementary Filters}
|
||||
\label{sec:org5b39592}
|
||||
\subsection{Synthesis of a set of three complementary filters}
|
||||
\label{sec:hinf_three_comp_filters}
|
||||
|
||||
Some applications may require to merge more than two sensors~\cite{stoten01_fusion_kinet_data_using_compos_filter,becker15_compl_filter_desig_three_frequen_bands}.
|
||||
For instance at the LIGO~\cite{matichard15_seism_isolat_advan_ligo}, three sensors (an LVDT, a seismometer and a geophone) are merged to form a super sensor (Fig.~\ref{fig:ligo_super_sensor_architecture}).\par
|
||||
For instance at the LIGO, three sensors (an LVDT, a seismometer and a geophone) are merged to form a super sensor~\cite{matichard15_seism_isolat_advan_ligo} (Fig.~\ref{fig:ligo_super_sensor_architecture}).\par
|
||||
|
||||
When merging \(n>2\) sensors using complementary filters, two architectures can be used as shown in Fig.~\ref{fig:sensor_fusion_three}.
|
||||
The fusion can either be done in a ``sequential'' way where \(n-1\) sets of two complementary filters are used (Fig.~\ref{fig:sensor_fusion_three_sequential}), or in a ``parallel'' way where one set of \(n\) complementary filters is used (Fig.~\ref{fig:sensor_fusion_three_parallel}).
|
||||
|
||||
In the first case, typical sensor fusion synthesis techniques can be used.
|
||||
However, when a parallel architecture is used, a new synthesis method for a set of more than two complementary filters is required as only simple analytical formulas have been proposed in the literature~\{\cite{stoten01_fusion_kinet_data_using_compos_filter,becker15_compl_filter_desig_three_frequen_bands}\}.
|
||||
However, when a parallel architecture is used, a new synthesis method for a set of more than two complementary filters is required as only simple analytical formulas have been proposed in the literature~\cite{stoten01_fusion_kinet_data_using_compos_filter,becker15_compl_filter_desig_three_frequen_bands}.
|
||||
A generalization of the proposed synthesis method of complementary filters is presented in this section. \par
|
||||
|
||||
\begin{figure}[htbp]
|
||||
\begin{subfigure}[b]{0.59\linewidth}
|
||||
\centering
|
||||
\includegraphics[scale=1]{figs/sensor_fusion_three_sequential.pdf}
|
||||
\caption{\label{fig:sensor_fusion_three_sequential}Sequential fusion}
|
||||
\caption{\label{fig:sensor_fusion_three_sequential}Sequential fusion.}
|
||||
\end{subfigure}
|
||||
\hfill
|
||||
\begin{subfigure}[b]{0.39\linewidth}
|
||||
\centering
|
||||
\includegraphics[scale=1]{figs/sensor_fusion_three_parallel.pdf}
|
||||
\caption{\label{fig:sensor_fusion_three_parallel}Parallel fusion}
|
||||
\caption{\label{fig:sensor_fusion_three_parallel}Parallel fusion.}
|
||||
\end{subfigure}
|
||||
\caption{\label{fig:sensor_fusion_three}Possible sensor fusion architecture when more than two sensors are to be merged}
|
||||
\caption{\label{fig:sensor_fusion_three}Possible sensor fusion architecture when more than two sensors are to be merged.}
|
||||
\centering
|
||||
\end{figure}
|
||||
|
||||
@@ -699,62 +683,16 @@ The synthesis objective is to compute a set of \(n\) stable transfer functions \
|
||||
\begin{subequations}
|
||||
\label{eq:hinf_problem_gen}
|
||||
\begin{align}
|
||||
& \sum_{i=0}^n H_i(s) = 1 \label{eq:hinf_cond_compl_gen} \\
|
||||
& \sum_{i=1}^n H_i(s) = 1 \label{eq:hinf_cond_compl_gen} \\
|
||||
& \left| H_i(j\omega) \right| < \frac{1}{\left| W_i(j\omega) \right|}, \quad \forall \omega,\ i = 1 \dots n \label{eq:hinf_cond_perf_gen}
|
||||
\end{align}
|
||||
\end{subequations}
|
||||
where \([W_1(s),\ W_2(s),\ \dots,\ W_n(s)]\) are weighting transfer functions that are chosen to specify the maximum wanted norms of the complementary filters during the synthesis.
|
||||
|
||||
Such synthesis objective is very close to the one described in Section~\ref{sec:synthesis_objective}, and indeed the proposed synthesis method is a generalization of the one presented in Section~\ref{sec:hinf_synthesis}. \par
|
||||
\([W_1(s),\ W_2(s),\ \dots,\ W_n(s)]\) are weighting transfer functions that are chosen to specify the maximum complementary filters' norm during the synthesis.
|
||||
|
||||
Before presenting the generalized synthesis method, the case with three sensors is presented.
|
||||
Consider the generalized plant \(P_3(s)\) shown in Fig.~\ref{fig:comp_filter_three_hinf_gen_plant} which is also described by~\eqref{eq:generalized_plant_three_filters}.
|
||||
Such synthesis objective is closely related to the one described in Section~\ref{sec:synthesis_objective}, and indeed the proposed synthesis method is a generalization of the one presented in Section~\ref{sec:hinf_synthesis}. \par
|
||||
|
||||
\begin{equation}
|
||||
\label{eq:generalized_plant_three_filters}
|
||||
\begin{bmatrix} z_1 \\ z_2 \\ z_3 \\ v \end{bmatrix} = P_3(s) \begin{bmatrix} w \\ u_1 \\ u_2 \end{bmatrix}; \quad P_3(s) = \begin{bmatrix}W_1(s) & -W_1(s) & -W_1(s) \\ 0 & \phantom{+}W_2(s) & 0 \\ 0 & 0 & \phantom{+}W_3(s) \\ 1 & 0 & 0 \end{bmatrix}
|
||||
\end{equation}
|
||||
|
||||
\begin{figure}[htbp]
|
||||
\begin{subfigure}[t]{0.5\linewidth}
|
||||
\centering
|
||||
\includegraphics[scale=1]{figs/comp_filter_three_hinf_gen_plant.pdf}
|
||||
\caption{\label{fig:comp_filter_three_hinf_gen_plant} Generalized plant}
|
||||
\vfill
|
||||
\end{subfigure}
|
||||
\hfill
|
||||
\begin{subfigure}[t]{0.5\linewidth}
|
||||
\centering
|
||||
\includegraphics[scale=1]{figs/comp_filter_three_hinf_fb.pdf}
|
||||
\caption{\label{fig:comp_filter_three_hinf_fb} Generalized plant with the synthesized filter}
|
||||
\end{subfigure}
|
||||
\hfill
|
||||
\caption{\label{fig:comp_filter_three_hinf} Architecture for the $\mathcal{H}_\infty$ synthesis of three complementary filters}
|
||||
\centering
|
||||
\end{figure}
|
||||
|
||||
Applying the \(\mathcal{H}_\infty\) synthesis on the generalized plant \(P_3(s)\) is equivalent as to find two stable filters \([H_2(s),\ H_3(s)]\) (shown in Fig.~\ref{fig:comp_filter_three_hinf_fb}) such that the \(\mathcal{H}_\infty\) norm of the transfer function from \(w\) to \([z_1,\ z_2, \ z_3]\) is less than one~\eqref{eq:hinf_syn_obj_three}.
|
||||
|
||||
\begin{equation}
|
||||
\label{eq:hinf_syn_obj_three}
|
||||
\left\| \begin{matrix} \left[1 - H_2(s) - H_3(s)\right] W_1(s) \\ H_2(s) W_2(s) \\ H_3(s) W_3(s) \end{matrix} \right\|_\infty \le 1
|
||||
\end{equation}
|
||||
|
||||
By defining \(H_1(s) \triangleq 1 - H_2(s) - H_3(s)\), the proposed \(\mathcal{H}_\infty\) synthesis solves the design problem \eqref{eq:hinf_problem_gen} with \(n=3\). \par
|
||||
|
||||
An example is given to validate the method where three sensors are used in different frequency bands.
|
||||
For instance a displacement sensor from DC up to \(\SI{1}{Hz}\), a geophone from \(1\) to \(\SI{10}{Hz}\) and an accelerometer above \(\SI{10}{Hz}\).
|
||||
Three weighting functions are designed using formula~\eqref{eq:weight_formula} and their inverse magnitudes are shown in Fig.~\ref{fig:three_complementary_filters_results} (dashed curves).
|
||||
The \(\mathcal{H}_\infty\) synthesis is performed on the generalized plant \(P_3(s)\) and the bode plot of the obtained complementary filters are shown in Fig.~\ref{fig:three_complementary_filters_results} (solid lines). \par
|
||||
|
||||
\begin{figure}[htbp]
|
||||
\centering
|
||||
\includegraphics[scale=1]{figs/three_complementary_filters_results.pdf}
|
||||
\caption{\label{fig:three_complementary_filters_results}Bode plot of the inverse weighting functions and of the three complementary filters obtained using the \(\mathcal{H}_\infty\) synthesis}
|
||||
\end{figure}
|
||||
|
||||
Even though there might not be any practical application for a set of more than 3 complementary filters, it can still be designed using the same procedure.
|
||||
A set of \(n\) complementary filters can be shaped using the generalized plant \(P_n(s)\) described by~\eqref{eq:generalized_plant_n_filters}.
|
||||
A set of \(n\) complementary filters can be shaped by applying the standard \(\mathcal{H}_\infty\) synthesis to the generalized plant \(P_n(s)\) described by~\eqref{eq:generalized_plant_n_filters}.
|
||||
|
||||
\begin{equation}
|
||||
\label{eq:generalized_plant_n_filters}
|
||||
@@ -769,28 +707,85 @@ A set of \(n\) complementary filters can be shaped using the generalized plant \
|
||||
\end{bmatrix}
|
||||
\end{equation}
|
||||
|
||||
If the synthesis if successful, a set of \(n-1\) filters \([H_2(s),\ H_3(s),\ \dots,\ H_n(s)]\) are obtained such that~\eqref{eq:hinf_syn_obj_gen} is verified.
|
||||
|
||||
\begin{equation}
|
||||
\label{eq:hinf_syn_obj_gen}
|
||||
\left\|\begin{matrix} \left(1 - \left[ H_2(s) + H_3(s) + \dots + H_n(s) \right]\right) W_1(s) \\ H_2(s) W_2(s) \\ \vdots \\ H_n(s) W_n(s) \end{matrix}\right\|_\infty \le 1
|
||||
\end{equation}
|
||||
|
||||
\(H_1(s)\) is then defined using~\eqref{eq:h1_comp_h2_hn} which is ensuring the complementary property for the set of \(n\) filters~\eqref{eq:hinf_cond_compl_gen}.
|
||||
Condition~\eqref{eq:hinf_cond_perf_gen} is satisfied thanks to~\eqref{eq:hinf_syn_obj_gen}.
|
||||
|
||||
\begin{equation}
|
||||
\label{eq:h1_comp_h2_hn}
|
||||
H_1(s) \triangleq 1 - \big[ H_2(s) + H_3(s) + \dots + H_n(s) \big]
|
||||
\end{equation}
|
||||
|
||||
An example is given to validate the proposed method for the synthesis of a set of three complementary filters.
|
||||
The sensors to be merged are a displacement sensor from DC up to \(\SI{1}{Hz}\), a geophone from \(1\) to \(\SI{10}{Hz}\) and an accelerometer above \(\SI{10}{Hz}\).
|
||||
Three weighting functions are designed using formula~\eqref{eq:weight_formula} and their inverse magnitude are shown in Fig.~\ref{fig:three_complementary_filters_results} (dashed curves).
|
||||
|
||||
Consider the generalized plant \(P_3(s)\) shown in Fig.~\ref{fig:comp_filter_three_hinf_gen_plant} which is also described by~\eqref{eq:generalized_plant_three_filters}.
|
||||
|
||||
\begin{equation}
|
||||
\label{eq:generalized_plant_three_filters}
|
||||
\begin{bmatrix} z_1 \\ z_2 \\ z_3 \\ v \end{bmatrix} = P_3(s) \begin{bmatrix} w \\ u_1 \\ u_2 \end{bmatrix}; \quad P_3(s) = \begin{bmatrix}W_1(s) & -W_1(s) & -W_1(s) \\ 0 & \phantom{+}W_2(s) & 0 \\ 0 & 0 & \phantom{+}W_3(s) \\ 1 & 0 & 0 \end{bmatrix}
|
||||
\end{equation}
|
||||
|
||||
\begin{figure}[htbp]
|
||||
\begin{subfigure}[t]{0.5\linewidth}
|
||||
\centering
|
||||
\includegraphics[scale=1]{figs/comp_filter_three_hinf_gen_plant.pdf}
|
||||
\caption{\label{fig:comp_filter_three_hinf_gen_plant} Generalized plant.}
|
||||
\vfill
|
||||
\end{subfigure}
|
||||
\hfill
|
||||
\begin{subfigure}[t]{0.5\linewidth}
|
||||
\centering
|
||||
\includegraphics[scale=1]{figs/comp_filter_three_hinf_fb.pdf}
|
||||
\caption{\label{fig:comp_filter_three_hinf_fb} Generalized plant with the synthesized filter.}
|
||||
\end{subfigure}
|
||||
\hfill
|
||||
\caption{\label{fig:comp_filter_three_hinf} Architecture for the $\mathcal{H}_\infty$ synthesis of three complementary filters.}
|
||||
\centering
|
||||
\end{figure}
|
||||
|
||||
The standard \(\mathcal{H}_\infty\) synthesis is performed on the generalized plant \(P_3(s)\).
|
||||
Two filters \(H_2(s)\) and \(H_3(s)\) are obtained such that the \(\mathcal{H}_\infty\) norm of the closed-loop transfer from \(w\) to \([z_1,\ z_2,\ z_3]\) of the system in Fig.~\ref{fig:comp_filter_three_hinf_fb} is less than one.
|
||||
Filter \(H_1(s)\) is defined using~\eqref{eq:h1_compl_h2_h3} thus ensuring the complementary property of the obtained set of filters.
|
||||
|
||||
\begin{equation}
|
||||
\label{eq:h1_compl_h2_h3}
|
||||
H_1(s) \triangleq 1 - \big[ H_2(s) + H_3(s) \big]
|
||||
\end{equation}
|
||||
|
||||
Figure~\ref{fig:three_complementary_filters_results} displays the three synthesized complementary filters (solid lines) which confirms that the synthesis is successful.\par
|
||||
|
||||
\begin{figure}[htbp]
|
||||
\centering
|
||||
\includegraphics[scale=1]{figs/three_complementary_filters_results.pdf}
|
||||
\caption{\label{fig:three_complementary_filters_results}Bode plot of the inverse weighting functions and of the three complementary filters obtained using the \(\mathcal{H}_\infty\) synthesis.}
|
||||
\end{figure}
|
||||
|
||||
\section{Conclusion}
|
||||
\label{sec:org369c466}
|
||||
\label{sec:conclusion}
|
||||
|
||||
Sensors measuring a physical quantities are always subject to limitations both in terms of bandwidth or accuracy.
|
||||
Complementary filters can be used to fuse multiple sensors with different characteristics in order to combine their benefits and yield a better estimate of the measured physical quantity.
|
||||
|
||||
The sensor fusion robustness and the obtained super sensor noise can be linked to the magnitude of the complementary filters.
|
||||
In this paper, a synthesis method that enables the shaping of the complementary filters norms has been proposed.
|
||||
Several example were used to emphasize the simplicity and the effectiveness of the proposed method.
|
||||
Links with ``closed-loop'' complementary filters where highlighted, and the proposed method was generalized for the design of a set of more than two complementary filters.
|
||||
A new method for designing complementary filters using the \(\mathcal{H}_\infty\) synthesis has been proposed.
|
||||
It allows to shape the magnitude of the filters by the use of weighting functions during the synthesis.
|
||||
This is very valuable in practice as the characteristics of the super sensor are linked to the complementary filters' magnitude.
|
||||
Therefore typical sensor fusion objectives can be translated into requirements on the magnitudes of the filters.
|
||||
Several examples were used to emphasize the simplicity and the effectiveness of the proposed method.
|
||||
|
||||
However, the shaping of the complementary filters' magnitude does not allow to directly optimize the super sensor noise and dynamical characteristics.
|
||||
Future work will aim at developing a complementary filter synthesis method that minimizes the super sensor noise while ensuring the robustness of the fusion.
|
||||
|
||||
\section*{Acknowledgment}
|
||||
\label{sec:orgd71260d}
|
||||
This research benefited from a FRIA grant from the French Community of Belgium.
|
||||
This paper is based on a paper previously presented at the ICCMA conference~\cite{dehaeze19_compl_filter_shapin_using_synth}.
|
||||
This paper has been assigned the LIGO document number LIGO-P2100328.
|
||||
|
||||
\section*{Data Availability}
|
||||
\label{sec:org30a4627}
|
||||
Matlab \cite{matlab20} was used for this study.
|
||||
Matlab~\cite{matlab20} was used for this study.
|
||||
The source code is available under a MIT License and archived in Zenodo~\cite{dehaeze21_new_method_desig_compl_filter_code}.
|
||||
|
||||
\bibliographystyle{elsarticle-num}
|
||||
|
||||
@@ -14,7 +14,7 @@
|
||||
doi = {10.1117/12.552518},
|
||||
school = {stanford university},
|
||||
title = {Low frequency vibration isolation and alignment system for
|
||||
advanced {LIGO}},
|
||||
{Advanced LIGO}},
|
||||
year = 2005,
|
||||
}
|
||||
|
||||
@@ -31,7 +31,7 @@
|
||||
@article{matichard15_seism_isolat_advan_ligo,
|
||||
author = {Matichard, F and Lantz, B and Mittleman, R and Mason, K and
|
||||
Kissel, J and others},
|
||||
title = {Seismic Isolation of Advanced {LIGO}: Review of Strategy,
|
||||
title = {Seismic Isolation of {Advanced LIGO}: Review of Strategy,
|
||||
Instrumentation and Performance},
|
||||
journal = {Classical and Quantum Gravity},
|
||||
volume = 32,
|
||||
@@ -285,7 +285,7 @@
|
||||
Giaime, Joseph A and Hammond, Giles Dominic and Hardham, C and
|
||||
Hennessy, Mike and How, Jonathan P and Lantz, Brian T and
|
||||
Macinnis, M and others},
|
||||
title = {Low-frequency active vibration isolation for advanced {LIGO}},
|
||||
title = {Low-frequency active vibration isolation for {Advanced LIGO}},
|
||||
booktitle = {Gravitational Wave and Particle Astrophysics Detectors},
|
||||
year = 2004,
|
||||
volume = 5500,
|
||||
@@ -303,10 +303,10 @@
|
||||
}
|
||||
|
||||
@phdthesis{lucia18_low_frequen_optim_perfor_advan,
|
||||
author = {Trozzo Lucia},
|
||||
author = {L. Trozzo},
|
||||
school = {University of Siena},
|
||||
title = {Low Frequency Optimization and Performance of Advanced
|
||||
Virgo Seismic Isolation System},
|
||||
title = {Low Frequency Optimization and Performance of {Advanced
|
||||
Virgo} Seismic Isolation System},
|
||||
year = 2018,
|
||||
}
|
||||
|
||||
@@ -378,11 +378,21 @@
|
||||
}
|
||||
|
||||
@misc{dehaeze21_new_method_desig_compl_filter_code,
|
||||
author = {Thomas Dehaeze},
|
||||
doi = {10.5281/zenodo.3894342},
|
||||
howpublished = {Source Code on Zonodo},
|
||||
month = 09,
|
||||
title = {A New Method of Designing Complementary Filters for Sensor
|
||||
Fusion using $\mathcal{H}_\infty$ Synthesis},
|
||||
author = {Dehaeze, Thomas},
|
||||
doi = {10.5281/zenodo.5361943},
|
||||
howpublished = {Source code on Zenodo},
|
||||
month = sep,
|
||||
title = {{A New Method of Designing Complementary Filters for Sensor
|
||||
Fusion Using the $\mathcal{H}_\infty$ Synthesis}},
|
||||
year = 2021,
|
||||
}
|
||||
|
||||
@book{skogestad07_multiv_feedb_contr,
|
||||
author = {Skogestad, Sigurd and Postlethwaite, Ian},
|
||||
title = {Multivariable Feedback Control: Analysis and Design -
|
||||
Second Edition},
|
||||
year = 2007,
|
||||
publisher = {John Wiley},
|
||||
isbn = {978-0470011683},
|
||||
note = {isbn:978-0470011683},
|
||||
}
|
||||
|
||||
98
matlab/.latexmkrc
Normal file
98
matlab/.latexmkrc
Normal file
@@ -0,0 +1,98 @@
|
||||
#!/bin/env perl
|
||||
|
||||
# Shebang is only to get syntax highlighting right across GitLab, GitHub and IDEs.
|
||||
# This file is not meant to be run, but read by `latexmk`.
|
||||
|
||||
# ======================================================================================
|
||||
# Perl `latexmk` configuration file
|
||||
# ======================================================================================
|
||||
|
||||
# ======================================================================================
|
||||
# PDF Generation/Building/Compilation
|
||||
# ======================================================================================
|
||||
|
||||
@default_files=('dehaeze21_desig_compl_filte_matlab.tex');
|
||||
|
||||
# PDF-generating modes are:
|
||||
# 1: pdflatex, as specified by $pdflatex variable (still largely in use)
|
||||
# 2: postscript conversion, as specified by the $ps2pdf variable (useless)
|
||||
# 3: dvi conversion, as specified by the $dvipdf variable (useless)
|
||||
# 4: lualatex, as specified by the $lualatex variable (best)
|
||||
# 5: xelatex, as specified by the $xelatex variable (second best)
|
||||
$pdf_mode = 1;
|
||||
|
||||
# Treat undefined references and citations as well as multiply defined references as
|
||||
# ERRORS instead of WARNINGS.
|
||||
# This is only checked in the *last* run, since naturally, there are undefined references
|
||||
# in initial runs.
|
||||
# This setting is potentially annoying when debugging/editing, but highly desirable
|
||||
# in the CI pipeline, where such a warning should result in a failed pipeline, since the
|
||||
# final document is incomplete/corrupted.
|
||||
#
|
||||
# However, I could not eradicate all warnings, so that `latexmk` currently fails with
|
||||
# this option enabled.
|
||||
# Specifically, `microtype` fails together with `fontawesome`/`fontawesome5`, see:
|
||||
# https://tex.stackexchange.com/a/547514/120853
|
||||
# The fix in that answer did not help.
|
||||
# Setting `verbose=silent` to mute `microtype` warnings did not work.
|
||||
# Switching between `fontawesome` and `fontawesome5` did not help.
|
||||
$warnings_as_errors = 0;
|
||||
|
||||
# Show used CPU time. Looks like: https://tex.stackexchange.com/a/312224/120853
|
||||
$show_time = 1;
|
||||
|
||||
# Default is 5; we seem to need more owed to the complexity of the document.
|
||||
# Actual documents probably don't need this many since they won't use all features,
|
||||
# plus won't be compiling from cold each time.
|
||||
$max_repeat=7;
|
||||
|
||||
# --shell-escape option (execution of code outside of latex) is required for the
|
||||
#'svg' package.
|
||||
# It converts raw SVG files to the PDF+PDF_TEX combo using InkScape.
|
||||
#
|
||||
# SyncTeX allows to jump between source (code) and output (PDF) in IDEs with support
|
||||
# (many have it). A value of `1` is enabled (gzipped), `-1` is enabled but uncompressed,
|
||||
# `0` is off.
|
||||
# Testing in VSCode w/ LaTeX Workshop only worked for the compressed version.
|
||||
# Adjust this as needed. Of course, only relevant for local use, no effect on a remote
|
||||
# CI pipeline (except for slower compilation, probably).
|
||||
#
|
||||
# %O and %S will forward Options and the Source file, respectively, given to latexmk.
|
||||
#
|
||||
# `set_tex_cmds` applies to all *latex commands (latex, xelatex, lualatex, ...), so
|
||||
# no need to specify these each. This allows to simply change `$pdf_mode` to get a
|
||||
# different engine. Check if this works with `latexmk --commands`.
|
||||
set_tex_cmds("--shell-escape -interaction=nonstopmode --synctex=1 %O %S");
|
||||
|
||||
# Use default pdf viewer
|
||||
$pdf_previewer = 'zathura';
|
||||
|
||||
# option 2 is same as 1 (run biber when necessary), but also deletes the
|
||||
# regeneratable bbl-file in a clenaup (`latexmk -c`). Do not use if original
|
||||
# bib file is not available!
|
||||
$bibtex_use = 2; # default: 1
|
||||
|
||||
# Change default `biber` call, help catch errors faster/clearer. See
|
||||
# https://web.archive.org/web/20200526101657/https://www.semipol.de/2018/06/12/latex-best-practices.html#database-entries
|
||||
$biber = "biber --validate-datamodel %O %S";
|
||||
|
||||
# ======================================================================================
|
||||
# Auxiliary Files
|
||||
# ======================================================================================
|
||||
|
||||
# Let latexmk know about generated files, so they can be used to detect if a
|
||||
# rerun is required, or be deleted in a cleanup.
|
||||
# loe: List of Examples (KOMAScript)
|
||||
# lol: List of Listings (`listings` and `minted` packages)
|
||||
# run.xml: biber runs
|
||||
# glg: glossaries log
|
||||
# glstex: generated from glossaries-extra
|
||||
push @generated_exts, 'loe', 'lol', 'run.xml', 'glg', 'glstex';
|
||||
|
||||
# Also delete the *.glstex files from package glossaries-extra. Problem is,
|
||||
# that that package generates files of the form "basename-digit.glstex" if
|
||||
# multiple glossaries are present. Latexmk looks for "basename.glstex" and so
|
||||
# does not find those. For that purpose, use wildcard.
|
||||
# Also delete files generated by gnuplot/pgfplots contour plots
|
||||
# (.dat, .script, .table).
|
||||
$clean_ext = "%R-*.glstex %R_contourtmp*.*";
|
||||
File diff suppressed because it is too large
Load Diff
@@ -1,13 +1,22 @@
|
||||
#+TITLE: A new method of designing complementary filters for sensor fusion using the $\mathcal{H}_\infty$ synthesis - Matlab Computation
|
||||
#+TITLE: Designing Complementary filters for sensor fusion using $\mathcal{H}_\infty$ synthesis - Matlab Computation
|
||||
:DRAWER:
|
||||
#+LANGUAGE: en
|
||||
#+EMAIL: dehaeze.thomas@gmail.com
|
||||
#+AUTHOR: Dehaeze Thomas
|
||||
|
||||
#+HTML_LINK_HOME: ../index.html
|
||||
#+HTML_LINK_UP: ../index.html
|
||||
|
||||
#+HTML_HEAD: <link rel="stylesheet" type="text/css" href="https://research.tdehaeze.xyz/css/style.css"/>
|
||||
#+HTML_HEAD: <script type="text/javascript" src="https://research.tdehaeze.xyz/js/script.js"></script>
|
||||
|
||||
#+LaTeX_CLASS: scrreprt
|
||||
#+LaTeX_HEADER_EXTRA: \input{/home/thomas/Cloud/org-theme/preamble.tex}
|
||||
#+BIND: org-latex-image-default-option "scale=1"
|
||||
#+BIND: org-latex-image-default-width ""
|
||||
|
||||
#+LATEX_CLASS: scrreprt
|
||||
#+LATEX_CLASS_OPTIONS: [a4paper, 10pt, DIV=12, parskip=full]
|
||||
#+LATEX_HEADER_EXTRA: \input{preamble.tex}
|
||||
#+LATEX_HEADER_EXTRA: \addbibresource{ref.bib}
|
||||
|
||||
#+PROPERTY: header-args:matlab :session *MATLAB*
|
||||
#+PROPERTY: header-args:matlab+ :comments none
|
||||
@@ -19,15 +28,27 @@
|
||||
#+PROPERTY: header-args:matlab+ :output-dir figs
|
||||
:END:
|
||||
|
||||
#+begin_export html
|
||||
<hr>
|
||||
<p>This report is also available as a <a href="./dehaeze21_desig_compl_filte_matlab.pdf">pdf</a>.</p>
|
||||
<hr>
|
||||
#+end_export
|
||||
|
||||
#+begin_export latex
|
||||
\clearpage
|
||||
#+end_export
|
||||
|
||||
* Introduction :ignore:
|
||||
|
||||
This file is the Matlab file for the paper cite:dehaeze21_new_method_desig_compl_filter.
|
||||
The present document is a companion file for the journal paper cite:dehaeze21_new_method_desig_compl_filter.
|
||||
All the Matlab cite:matlab20 scripts used in the paper are here shared and explained.
|
||||
|
||||
This document is divided into several sections:
|
||||
- in section [[#sec:h_inf_synthesis_complementary_filters]], the $\mathcal{H}_\infty$ synthesis is used for generating two complementary filters
|
||||
- in section [[#sec:three_comp_filters]], a method using the $\mathcal{H}_\infty$ synthesis is proposed to shape three of more complementary filters
|
||||
- in section [[#sec:comp_filters_ligo]], the $\mathcal{H}_\infty$ synthesis is used and compared with FIR complementary filters used for LIGO
|
||||
- in section [[#sec:closed_loop_complementary_filters]]
|
||||
This document is divided into the following sections also corresponding to the paper sections:
|
||||
- Section [[#sec:h_inf_synthesis_complementary_filters]]: the shaping of complementary filters is written as an $\mathcal{H}_\infty$ optimization problem using weighting functions. The weighting function design is discussed and the method is applied for the design of a set of simple complementary filters.
|
||||
- Section [[#sec:comp_filters_ligo]]: the effectiveness of the proposed complementary filter synthesis strategy is demonstrated by designing complex complementary filters used in the first isolation stage at the LIGO
|
||||
- Section [[#sec:closed_loop_complementary_filters]]: complementary filters are designed using the classical feedback loop
|
||||
- Section [[#sec:three_comp_filters]]: the proposed design method is generalized for the design of a set of three complementary filters
|
||||
- Section [[#sec:matlab_scripts]]: complete Matlab scripts and functions developed are listed
|
||||
|
||||
* H-Infinity synthesis of complementary filters
|
||||
:PROPERTIES:
|
||||
@@ -35,10 +56,46 @@ This document is divided into several sections:
|
||||
:CUSTOM_ID: sec:h_inf_synthesis_complementary_filters
|
||||
:END:
|
||||
|
||||
** Introduction :ignore:
|
||||
#+begin_note
|
||||
The Matlab file corresponding to this section is accessible [[file:matlab/1_synthesis_complementary_filters.m][here]].
|
||||
#+end_note
|
||||
** Synthesis Architecture
|
||||
In order to generate two complementary filters with a wanted shape, the generalized plant of Figure [[fig:h_infinity_robust_fusion_plant]] can be used.
|
||||
The included weights $W_1(s)$ and $W_2(s)$ are used to specify the upper bounds of the complementary filters being generated.
|
||||
|
||||
#+name: fig:h_infinity_robust_fusion_plant
|
||||
#+caption: Generalized plant used for the $\mathcal{H}_\infty$ synthesis of a set of two complementary fiters
|
||||
[[file:figs-journal/h_infinity_robust_fusion_plant.png]]
|
||||
|
||||
Applied the standard $\mathcal{H}_\infty$ synthesis on this generalized plant will give a transfer function $H_2(s)$ (see Figure [[fig:h_infinity_robust_fusion_fb]]) such that the $\mathcal{H}_\infty$ norm of the transfer function from $w$ to $[z_1,\ z_2]$ is less than one eqref:eq:h_inf_objective.
|
||||
|
||||
#+name: fig:h_infinity_robust_fusion_fb
|
||||
#+caption: Generalized plant with the synthesized filter obtained after the $\mathcal{H}_\infty$ synthesis
|
||||
[[file:figs-journal/h_infinity_robust_fusion_fb.png]]
|
||||
|
||||
\begin{equation}
|
||||
\left\| \begin{array}{c} (1 - H_2(s)) W_1(s) \\ H_2(s) W_2(s) \end{array} \right\|_\infty < 1 \label{eq:h_inf_objective}
|
||||
\end{equation}
|
||||
|
||||
Thus, if the synthesis is successful and the above condition is verified, we can define $H_1(s)$ to be the complementary of $H_2(s)$ eqref:eq:H1_complementary_of_H2 and we have condition eqref:eq:shaping_comp_filters verified.
|
||||
|
||||
\begin{equation}
|
||||
H_1(s) = 1 - H_2(s) \label{eq:H1_complementary_of_H2}
|
||||
\end{equation}
|
||||
|
||||
\begin{equation}
|
||||
\left\| \begin{array}{c} H_1(s) W_1(s) \\ H_2(s) W_2(s) \end{array} \right\|_\infty < 1
|
||||
\quad \Longrightarrow \quad
|
||||
\left\{
|
||||
\begin{array}{c}
|
||||
|H_1(j\omega)| < \frac{1}{|W_1(j\omega)|}, \quad \forall \omega \\
|
||||
|H_2(j\omega)| < \frac{1}{|W_2(j\omega)|}, \quad \forall \omega
|
||||
\end{array}
|
||||
\right. \label{eq:shaping_comp_filters}
|
||||
\end{equation}
|
||||
|
||||
We then see that $W_1(s)$ and $W_2(s)$ can be used to set the wanted upper bounds of the magnitudes of both $H_1(s)$ and $H_2(s)$.
|
||||
|
||||
The presented synthesis method therefore allows to shape two filters $H_1(s)$ and $H_2(s)$ eqref:eq:shaping_comp_filters while ensuring their complementary property eqref:eq:H1_complementary_of_H2.
|
||||
|
||||
The complete Matlab script for this part is given in Section [[#sec:1_synthesis_complementary_filters]].
|
||||
|
||||
** Matlab Init :noexport:ignore:
|
||||
#+begin_src matlab :tangle no :exports none :results silent :noweb yes :var current_dir=(file-name-directory buffer-file-name)
|
||||
@@ -56,45 +113,16 @@ freqs = logspace(-1, 3, 1000);
|
||||
|
||||
#+begin_src matlab :tangle no
|
||||
addpath('./matlab');
|
||||
addpath('./matlab/src');
|
||||
#+end_src
|
||||
|
||||
#+begin_src matlab :eval no
|
||||
%% Add functions to path
|
||||
addpath('./src');
|
||||
#+end_src
|
||||
|
||||
** Synthesis Architecture
|
||||
We here synthesize two complementary filters using the $\mathcal{H}_\infty$ synthesis.
|
||||
The goal is to specify upper bounds on the norms of the two complementary filters $H_1(s)$ and $H_2(s)$ while ensuring their complementary property ($H_1(s) + H_2(s) = 1$).
|
||||
|
||||
In order to do so, we use the generalized plant shown on figure [[fig:h_infinity_robst_fusion]] where $W_1(s)$ and $W_2(s)$ are weighting transfer functions that will be used to shape $H_1(s)$ and $H_2(s)$ respectively.
|
||||
|
||||
#+name: fig:h_infinity_robst_fusion
|
||||
#+caption: $\mathcal{H}_\infty$ synthesis of the complementary filters
|
||||
[[file:figs-journal/h_infinity_robust_fusion_plant.png]]
|
||||
|
||||
The $\mathcal{H}_\infty$ synthesis applied on this generalized plant will give a transfer function $H_2$ (figure [[fig:h_infinity_robst_fusion]]) such that the $\mathcal{H}_\infty$ norm of the transfer function from $w$ to $[z_1,\ z_2]$ is less than one:
|
||||
\[ \left\| \begin{array}{c} (1 - H_2(s)) W_1(s) \\ H_2(s) W_2(s) \end{array} \right\|_\infty < 1 \]
|
||||
|
||||
Thus, if the above condition is verified, we can define $H_1(s) = 1 - H_2(s)$ and we have that:
|
||||
\[ \left\| \begin{array}{c} H_1(s) W_1(s) \\ H_2(s) W_2(s) \end{array} \right\|_\infty < 1 \]
|
||||
Which is almost (with an maximum error of $\sqrt{2}$) equivalent to:
|
||||
\begin{align*}
|
||||
|H_1(j\omega)| &< \frac{1}{|W_1(j\omega)|}, \quad \forall \omega \\
|
||||
|H_2(j\omega)| &< \frac{1}{|W_2(j\omega)|}, \quad \forall \omega
|
||||
\end{align*}
|
||||
|
||||
We then see that $W_1(s)$ and $W_2(s)$ can be used to shape both $H_1(s)$ and $H_2(s)$ while ensuring their complementary property by the definition of $H_1(s) = 1 - H_2(s)$.
|
||||
|
||||
** Design of Weighting Function
|
||||
** Design of Weighting Function - Proposed formula
|
||||
A formula is proposed to help the design of the weighting functions:
|
||||
\begin{equation}
|
||||
W(s) = \left( \frac{
|
||||
\frac{1}{\omega_0} \sqrt{\frac{1 - \left(\frac{G_0}{G_c}\right)^{\frac{2}{n}}}{1 - \left(\frac{G_c}{G_\infty}\right)^{\frac{2}{n}}}} s + \left(\frac{G_0}{G_c}\right)^{\frac{1}{n}}
|
||||
}{
|
||||
\left(\frac{1}{G_\infty}\right)^{\frac{1}{n}} \frac{1}{\omega_0} \sqrt{\frac{1 - \left(\frac{G_0}{G_c}\right)^{\frac{2}{n}}}{1 - \left(\frac{G_c}{G_\infty}\right)^{\frac{2}{n}}}} s + \left(\frac{1}{G_c}\right)^{\frac{1}{n}}
|
||||
}\right)^n
|
||||
}\right)^n \label{eq:weighting_function_formula}
|
||||
\end{equation}
|
||||
|
||||
The parameters permits to specify:
|
||||
@@ -152,11 +180,16 @@ exportFig('figs/weight_formula.pdf', 'width', 'wide', 'height', 'normal');
|
||||
#+end_src
|
||||
|
||||
#+name: fig:weight_formula
|
||||
#+caption: Gain of the Weighting Function formula
|
||||
#+caption: Magnitude of the weighting function generated using formula eqref:eq:weighting_function_formula
|
||||
#+RESULTS:
|
||||
[[file:figs/weight_formula.png]]
|
||||
|
||||
** Example
|
||||
** Weighting functions for the design of two complementary filters
|
||||
<<sec:weighting_functions_example>>
|
||||
|
||||
The weighting function formula eqref:eq:weighting_function_formula is used to generate the upper bounds of two complementary filters that we wish to design.
|
||||
|
||||
The matlab function =generateWF= is described in Section [[#sec:generateWF]].
|
||||
|
||||
#+begin_src matlab
|
||||
%% Design of the Weighting Functions
|
||||
@@ -164,6 +197,8 @@ W1 = generateWF('n', 3, 'w0', 2*pi*10, 'G0', 1000, 'Ginf', 1/10, 'Gc', 0.45);
|
||||
W2 = generateWF('n', 2, 'w0', 2*pi*10, 'G0', 1/10, 'Ginf', 1000, 'Gc', 0.45);
|
||||
#+end_src
|
||||
|
||||
The inverse magnitude of these two weighting functions are shown in Figure [[fig:weights_W1_W2]].
|
||||
|
||||
#+begin_src matlab :exports none
|
||||
%% Plot of the Weighting function magnitude
|
||||
figure;
|
||||
@@ -196,12 +231,12 @@ exportFig('figs/weights_W1_W2.pdf', 'width', 'wide', 'height', 'normal');
|
||||
#+end_src
|
||||
|
||||
#+name: fig:weights_W1_W2
|
||||
#+caption: Weights on the complementary filters $W_1$ and $W_2$ and the associated performance weights
|
||||
#+caption: Inverse magnitude of the design weighting functions
|
||||
#+RESULTS:
|
||||
[[file:figs/weights_W1_W2.png]]
|
||||
|
||||
** H-Infinity Synthesis
|
||||
We define the generalized plant $P$ on matlab.
|
||||
** Synthesis of the complementary filters
|
||||
The generalized plant of Figure [[fig:h_infinity_robust_fusion_plant]] is defined as follows:
|
||||
#+begin_src matlab
|
||||
%% Generalized Plant
|
||||
P = [W1 -W1;
|
||||
@@ -209,7 +244,7 @@ P = [W1 -W1;
|
||||
1 0];
|
||||
#+end_src
|
||||
|
||||
And we do the $\mathcal{H}_\infty$ synthesis using the =hinfsyn= command.
|
||||
And the $\mathcal{H}_\infty$ synthesis is performed using the =hinfsyn= command.
|
||||
#+begin_src matlab :results output replace :exports both
|
||||
%% H-Infinity Synthesis
|
||||
[H2, ~, gamma, ~] = hinfsyn(P, 1, 1,'TOLGAM', 0.001, 'METHOD', 'ric', 'DISPLAY', 'on');
|
||||
@@ -242,20 +277,23 @@ And we do the $\mathcal{H}_\infty$ synthesis using the =hinfsyn= command.
|
||||
Best performance (actual): 0.9897
|
||||
#+end_example
|
||||
|
||||
We then define the high pass filter $H_1 = 1 - H_2$. The bode plot of both $H_1$ and $H_2$ is shown on figure [[fig:hinf_filters_results]].
|
||||
As shown above, the obtained $\mathcal{H}_\infty$ norm of the transfer function from $w$ to $[z_1,\ z_2]$ is found to be less than one meaning the synthesis is successful.
|
||||
|
||||
We then define the filter $H_1(s)$ to be the complementary of $H_2(s)$ eqref:eq:H1_complementary_of_H2.
|
||||
#+begin_src matlab
|
||||
%% Define H1 to be the complementary of H2
|
||||
H1 = 1 - H2;
|
||||
#+end_src
|
||||
|
||||
Or one can just used to =generateCF= Matlab function:
|
||||
The function =generateCF= can also be used to synthesize the complementary filters.
|
||||
This function is described in Section [[#sec:generateCF]].
|
||||
#+begin_src matlab :tangle no :eval no
|
||||
[H1, H2] = generateCF(W1, W2);
|
||||
#+end_src
|
||||
|
||||
** Obtained Complementary Filters
|
||||
The obtained complementary filters are shown on figure [[fig:hinf_filters_results]].
|
||||
The obtained complementary filters are shown below and are found to be of order 5.
|
||||
Their bode plots are shown in figure [[fig:hinf_filters_results]] and compare with the defined upper bounds.
|
||||
|
||||
#+begin_src matlab :results output replace :exports results :tangle no
|
||||
zpk(H1)
|
||||
@@ -343,13 +381,9 @@ exportFig('figs/hinf_filters_results.pdf', 'width', 'wide', 'height', 'tall');
|
||||
:END:
|
||||
|
||||
** Introduction :ignore:
|
||||
#+begin_note
|
||||
The Matlab file corresponding to this section is accessible [[file:matlab/2_ligo_complementary_filters.m][here]].
|
||||
#+end_note
|
||||
In this section, the proposed method for the design of complementary filters is validated for the design of a set of two complex complementary filters used for the first isolation stage at the LIGO cite:hua04_low_ligo.
|
||||
|
||||
Let's try to design complementary filters that are corresponding to the complementary filters design for the LIGO and described in cite:hua05_low_ligo.
|
||||
|
||||
The FIR complementary filters designed in cite:hua05_low_ligo are of order 512.
|
||||
The complete Matlab script for this part is given in Section [[#sec:2_ligo_complementary_filters]].
|
||||
|
||||
** Matlab Init :noexport:ignore:
|
||||
#+begin_src matlab :tangle no :exports none :results silent :noweb yes :var current_dir=(file-name-directory buffer-file-name)
|
||||
@@ -367,22 +401,16 @@ freqs = logspace(-3, 0, 1000);
|
||||
|
||||
#+begin_src matlab :tangle no
|
||||
addpath('./matlab');
|
||||
addpath('./matlab/src');
|
||||
#+end_src
|
||||
|
||||
#+begin_src matlab :eval no
|
||||
%% Add functions to path
|
||||
addpath('./src');
|
||||
#+end_src
|
||||
|
||||
** Specifications
|
||||
The specifications for the filters are:
|
||||
1. From $0$ to $0.008\text{ Hz}$,the magnitude of the filter’s transfer function should be less than or equal to $8 \times 10^{-3}$
|
||||
2. From $0.008\text{ Hz}$ to $0.04\text{ Hz}$, it attenuates the input signal proportional to frequency cubed
|
||||
3. Between $0.04\text{ Hz}$ and $0.1\text{ Hz}$, the magnitude of the transfer function should be less than 3
|
||||
4. Above $0.1\text{ Hz}$, the maximum of the magnitude of the complement filter should be as close to zero as possible. In our system, we would like to have the magnitude of the complementary filter to be less than $0.1$. As the filters obtained in cite:hua05_low_ligo have a magnitude of $0.045$, we will set that as our requirement
|
||||
The specifications for the set of complementary filters ($L_1,H_1$) used at the LIGO are summarized below (for further details, refer to cite:hua04_polyp_fir_compl_filter_contr_system):
|
||||
- From 0 to 0.008 Hz, the magnitude $|L_1(j\omega)|$ should be less or equal to $8 \times 10^{-4}$
|
||||
- Between 0.008 Hz to 0.04 Hz, the filter $L_1(s)$ should attenuate the input signal proportional to frequency cubed
|
||||
- Between 0.04 Hz to 0.1 Hz, the magnitude $|L_1(j\omega)|$ should be less than $3$
|
||||
- Above 0.1 Hz, the magnitude $|H_1(j\omega)|$ should be less than $0.045$
|
||||
|
||||
The specifications are translated in upper bounds of the complementary filters are shown on figure [[fig:ligo_specifications]].
|
||||
The specifications are translated into upper bounds of the complementary filters and are shown in Figure [[fig:ligo_specifications]].
|
||||
|
||||
#+begin_src matlab :exports none
|
||||
%% Upper bounds for the complementary filters
|
||||
@@ -415,10 +443,9 @@ exportFig('figs/ligo_specifications.pdf', 'width', 'wide', 'height', 'normal');
|
||||
[[file:figs/ligo_specifications.png]]
|
||||
|
||||
** FIR Filter
|
||||
We here try to implement the FIR complementary filter synthesis as explained in cite:hua05_low_ligo.
|
||||
For that, we use the [[http://cvxr.com/cvx/][CVX matlab Toolbox]].
|
||||
To replicated the complementary filters developed in cite:hua04_low_ligo, the CVX Matlab toolbox cite:grant14_cvx is used.
|
||||
|
||||
We setup the CVX toolbox and use the =SeDuMi= solver.
|
||||
The CVX toolbox is initialized and the =SeDuMi= solver cite:sturm99_using_sedum is used.
|
||||
#+begin_src matlab
|
||||
%% Initialized CVX
|
||||
cvx_startup;
|
||||
@@ -434,7 +461,7 @@ w3 = 0.04:8.12e-4:0.1;
|
||||
w4 = 0.1:8.12e-4:0.83;
|
||||
#+end_src
|
||||
|
||||
We then define the order of the FIR filter.
|
||||
The order $n$ of the FIR filter is defined.
|
||||
#+begin_src matlab
|
||||
%% Filter order
|
||||
n = 512;
|
||||
@@ -453,7 +480,7 @@ B3 = [zeros(length(w3),1), sin(kron(w3'.*(2*pi),[1:n-1]))];
|
||||
B4 = [zeros(length(w4),1), sin(kron(w4'.*(2*pi),[1:n-1]))];
|
||||
#+end_src
|
||||
|
||||
We run the convex optimization.
|
||||
And the convex optimization is run.
|
||||
#+begin_src matlab :results output replace :wrap example
|
||||
%% Convex optimization
|
||||
cvx_begin
|
||||
@@ -561,7 +588,7 @@ Optimal value (cvx_optval): -0.0433341
|
||||
h = y(2:end);
|
||||
#+end_example
|
||||
|
||||
Finally, we compute the filter response over the frequency vector defined and the result is shown on figure [[fig:fir_filter_ligo]] which is very close to the filters obtain in cite:hua05_low_ligo.
|
||||
Finally, the filter response is computed over the frequency vector defined and the result is shown on figure [[fig:fir_filter_ligo]] which is very close to the filters obtain in cite:hua04_low_ligo.
|
||||
|
||||
#+begin_src matlab
|
||||
%% Combine the frequency vectors to form the obtained filter
|
||||
@@ -621,17 +648,18 @@ exportFig('figs/fir_filter_ligo.pdf', 'width', 'wide', 'height', 'tall');
|
||||
#+RESULTS:
|
||||
[[file:figs/fir_filter_ligo.png]]
|
||||
|
||||
** Weights
|
||||
We design weights that will be used for the $\mathcal{H}_\infty$ synthesis of the complementary filters.
|
||||
** Weighting function design
|
||||
The weightings function that will be used for the $\mathcal{H}_\infty$ synthesis of the complementary filters are now designed.
|
||||
|
||||
These weights will determine the order of the obtained filters.
|
||||
|
||||
Here are the requirements on the filters:
|
||||
- reasonable order
|
||||
- to be as close as possible to the specified upper bounds
|
||||
- stable minimum phase
|
||||
- stable and minimum phase
|
||||
|
||||
The bode plot of the weights is shown on figure [[fig:ligo_weights]].
|
||||
|
||||
#+begin_src matlab :exports none
|
||||
The weighting function for the High Pass filter is defined as follows:
|
||||
#+begin_src matlab
|
||||
%% Design of the weight for the high pass filter
|
||||
w1 = 2*pi*0.008; x1 = 0.35;
|
||||
w2 = 2*pi*0.04; x2 = 0.5;
|
||||
@@ -651,7 +679,8 @@ wH = 1/wH;
|
||||
wH = minreal(ss(wH));
|
||||
#+end_src
|
||||
|
||||
#+begin_src matlab :exports none
|
||||
And the weighting function for the Low pass filter is taken as a Chebyshev Type I filter.
|
||||
#+begin_src matlab
|
||||
%% Design of the weight for the low pass filter
|
||||
n = 20; % Filter order
|
||||
Rp = 1; % Peak to peak passband ripple
|
||||
@@ -665,6 +694,8 @@ wL = 1/wL;
|
||||
wL = minreal(ss(wL));
|
||||
#+end_src
|
||||
|
||||
The inverse magnitude of the weighting functions are shown in Figure [[fig:ligo_weights]].
|
||||
|
||||
#+begin_src matlab :exports none
|
||||
%% Magnitude of the designed Weights and initial specifications
|
||||
figure;
|
||||
@@ -701,8 +732,8 @@ exportFig('figs/ligo_weights.pdf', 'width', 'wide', 'height', 'normal');
|
||||
#+RESULTS:
|
||||
[[file:figs/ligo_weights.png]]
|
||||
|
||||
** H-Infinity Synthesis
|
||||
We define the generalized plant as shown on figure [[fig:h_infinity_robst_fusion]].
|
||||
** Synthesis of the complementary filters
|
||||
The generalized plant of figure [[fig:h_infinity_robust_fusion_plant]] is defined.
|
||||
#+begin_src matlab
|
||||
%% Generalized plant for the H-infinity Synthesis
|
||||
P = [0 wL;
|
||||
@@ -710,7 +741,7 @@ P = [0 wL;
|
||||
1 0];
|
||||
#+end_src
|
||||
|
||||
And we do the $\mathcal{H}_\infty$ synthesis using the =hinfsyn= command.
|
||||
And the standard $\mathcal{H}_\infty$ synthesis using the =hinfsyn= command is performed.
|
||||
#+begin_src matlab :results output replace :exports both :wrap example
|
||||
%% Standard H-Infinity synthesis
|
||||
[Hl, ~, gamma, ~] = hinfsyn(P, 1, 1,'TOLGAM', 0.001, 'METHOD', 'ric', 'DISPLAY', 'on');
|
||||
@@ -740,7 +771,13 @@ Test bounds: 0.3276 < gamma <= 1.8063
|
||||
Gamma value achieved: 1.0439
|
||||
#+end_example
|
||||
|
||||
The high pass filter is defined as $H_H = 1 - H_L$.
|
||||
The obtained $\mathcal{H}_\infty$ norm is found to be close than one meaning the synthesis is successful.
|
||||
|
||||
The high pass filter $H_H(s)$ is defined to be the complementary of the synthesized low pass filter $H_L(s)$:
|
||||
\begin{equation}
|
||||
H_H(s) = 1 - H_L(s)
|
||||
\end{equation}
|
||||
|
||||
#+begin_src matlab
|
||||
%% High pass filter as the complementary of the low pass filter
|
||||
Hh = 1 - Hl;
|
||||
@@ -752,7 +789,7 @@ Hh = minreal(Hh);
|
||||
Hl = minreal(Hl);
|
||||
#+end_src
|
||||
|
||||
The size of the filters is shown below.
|
||||
The size of the filters is shown to be equal to the sum of the weighting functions orders.
|
||||
|
||||
#+begin_src matlab :exports results :results output replace :wrap example :tangle no
|
||||
size(Hh), size(Hl)
|
||||
@@ -765,7 +802,7 @@ State-space model with 1 outputs, 1 inputs, and 27 states.
|
||||
State-space model with 1 outputs, 1 inputs, and 27 states.
|
||||
#+end_example
|
||||
|
||||
The bode plot of the obtained filters as shown on figure [[fig:hinf_synthesis_ligo_results]].
|
||||
The magnitude of the obtained filters as well as the requirements are shown in Figure [[fig:hinf_synthesis_ligo_results]].
|
||||
|
||||
#+begin_src matlab :exports none
|
||||
%% Bode plot of the obtained filters and comparison with the upper bounds
|
||||
@@ -799,8 +836,11 @@ exportFig('figs/hinf_synthesis_ligo_results.pdf', 'width', 'wide', 'height', 'no
|
||||
#+RESULTS:
|
||||
[[file:figs/hinf_synthesis_ligo_results.png]]
|
||||
|
||||
** Compare FIR and H-Infinity Filters
|
||||
Let's now compare the FIR filters designed in cite:hua05_low_ligo and the one obtained with the $\mathcal{H}_\infty$ synthesis on figure [[fig:comp_fir_ligo_hinf]].
|
||||
** Comparison of the FIR filters and synthesized filters
|
||||
|
||||
Let's now compare the FIR filters designed in cite:hua04_low_ligo with the with complementary filters obtained with the $\mathcal{H}_\infty$ synthesis.
|
||||
|
||||
This is done in Figure [[fig:comp_fir_ligo_hinf]], and both set of filters are found to be very close to each other.
|
||||
|
||||
#+begin_src matlab :exports none
|
||||
%% Comparison of the complementary filters obtained with H-infinity and with CVX
|
||||
@@ -872,9 +912,14 @@ exportFig('figs/comp_fir_ligo_hinf.pdf', 'width', 'wide', 'height', 'tall');
|
||||
:END:
|
||||
|
||||
** Introduction :ignore:
|
||||
#+begin_note
|
||||
The Matlab file corresponding to this section is accessible [[file:matlab/3_closed_loop_complementary_filters.m][here]].
|
||||
#+end_note
|
||||
|
||||
In this section, the classical feedback architecture shown in Figure [[fig:feedback_sensor_fusion]] is used for the design of complementary filters.
|
||||
|
||||
#+name: fig:feedback_sensor_fusion
|
||||
#+caption: "Closed-Loop" complementary filters
|
||||
[[file:figs-journal/feedback_sensor_fusion.png]]
|
||||
|
||||
The complete Matlab script for this part is given in Section [[#sec:3_closed_loop_complementary_filters]].
|
||||
|
||||
** Matlab Init :noexport:ignore:
|
||||
#+begin_src matlab :tangle no :exports none :results silent :noweb yes :var current_dir=(file-name-directory buffer-file-name)
|
||||
@@ -892,34 +937,64 @@ freqs = logspace(-1, 3, 1000);
|
||||
|
||||
#+begin_src matlab :tangle no
|
||||
addpath('./matlab');
|
||||
addpath('./matlab/src');
|
||||
#+end_src
|
||||
|
||||
#+begin_src matlab :eval no
|
||||
%% Add functions to path
|
||||
addpath('./src');
|
||||
#+end_src
|
||||
** Weighting Function design
|
||||
Weighting functions using the =generateWF= Matlab function are designed to specify the upper bounds of the complementary filters to be designed.
|
||||
These weighting functions are the same as the ones used in Section [[sec:weighting_functions_example]].
|
||||
|
||||
** Using Feedback architecture
|
||||
#+begin_src matlab
|
||||
%% Design of the Weighting Functions
|
||||
W1 = generateWF('n', 3, 'w0', 2*pi*10, 'G0', 1000, 'Ginf', 1/10, 'Gc', 0.45);
|
||||
W2 = generateWF('n', 2, 'w0', 2*pi*10, 'G0', 1/10, 'Ginf', 1000, 'Gc', 0.45);
|
||||
#+end_src
|
||||
|
||||
Let's first synthesize $H_1(s)$:
|
||||
** Generalized plant
|
||||
The generalized plant of Figure [[fig:feedback_synthesis_architecture_generalized_plant]] is defined below:
|
||||
#+begin_src matlab
|
||||
%% Generalized plant for "closed-loop" complementary filter synthesis
|
||||
P = [ W1 0 1;
|
||||
-W1 W2 -1];
|
||||
#+end_src
|
||||
|
||||
#+name: fig:feedback_synthesis_architecture_generalized_plant
|
||||
#+caption: Generalized plant used for the $\mathcal{H}_\infty$ synthesis of "closed-loop" complementary filters
|
||||
[[file:figs-journal/feedback_synthesis_architecture_generalized_plant.png]]
|
||||
|
||||
** Synthesis of the closed-loop complementary filters
|
||||
And the standard $\mathcal{H}_\infty$ synthesis is performed.
|
||||
#+begin_src matlab :results output replace :exports both
|
||||
%% Standard H-Infinity Synthesis
|
||||
[L, ~, gamma, ~] = hinfsyn(P, 1, 1,'TOLGAM', 0.001, 'METHOD', 'ric', 'DISPLAY', 'on');
|
||||
#+end_src
|
||||
|
||||
#+RESULTS:
|
||||
#+begin_example
|
||||
Test bounds: 0.3191 <= gamma <= 1.669
|
||||
|
||||
gamma X>=0 Y>=0 rho(XY)<1 p/f
|
||||
7.299e-01 -1.5e-19 -2.4e+01 # 1.555e-18 f
|
||||
1.104e+00 0.0e+00 1.6e-07 2.037e-19 p
|
||||
8.976e-01 -3.2e-16 -1.4e+02 # 5.561e-16 f
|
||||
9.954e-01 0.0e+00 1.6e-07 1.041e-15 p
|
||||
9.452e-01 -1.1e-15 -3.8e+02 # 4.267e-15 f
|
||||
9.700e-01 -6.5e-16 -1.6e+03 # 9.876e-15 f
|
||||
9.826e-01 0.0e+00 1.6e-07 8.775e-39 p
|
||||
9.763e-01 -5.0e-16 -6.2e+03 # 3.519e-14 f
|
||||
9.795e-01 0.0e+00 1.6e-07 6.971e-20 p
|
||||
9.779e-01 -1.9e-31 -2.2e+04 # 5.600e-18 f
|
||||
9.787e-01 0.0e+00 1.6e-07 5.546e-19 p
|
||||
Limiting gains...
|
||||
9.789e-01 0.0e+00 1.6e-07 1.084e-13 p
|
||||
9.789e-01 0.0e+00 9.7e-07 1.137e-13 p
|
||||
|
||||
Best performance (actual): 0.9789
|
||||
#+end_example
|
||||
|
||||
** Synthesized filters
|
||||
The obtained filter $L(s)$ can then be included in the feedback architecture shown in Figure [[fig:hinf_filters_results_mixed_sensitivity]].
|
||||
|
||||
The closed-loop transfer functions from $\hat{x}_1$ to $\hat{x}$ and from $\hat{x}_2$ to $\hat{x}$ corresponding respectively to the sensitivity and complementary sensitivity transfer functions are defined below:
|
||||
#+begin_src matlab
|
||||
%% Complementary filters
|
||||
H1 = inv(1 + L);
|
||||
@@ -944,6 +1019,8 @@ zpk(H2) =
|
||||
(s+1.29e05) (s^2 + 102.1s + 2733) (s^2 + 69.45s + 3272)
|
||||
#+end_example
|
||||
|
||||
The bode plots of the synthesized complementary filters are compared with the upper bounds in Figure [[fig:hinf_filters_results_mixed_sensitivity]].
|
||||
|
||||
#+begin_src matlab :exports none
|
||||
%% Bode plot of the obtained Complementary filters with upper-bounds
|
||||
freqs = logspace(-1, 3, 1000);
|
||||
@@ -1000,21 +1077,21 @@ exportFig('figs/hinf_filters_results_mixed_sensitivity.pdf', 'width', 'wide', 'h
|
||||
#+end_src
|
||||
|
||||
#+name: fig:hinf_filters_results_mixed_sensitivity
|
||||
#+caption:
|
||||
#+caption: Bode plot of the obtained complementary filters
|
||||
#+RESULTS:
|
||||
[[file:figs/hinf_filters_results_mixed_sensitivity.png]]
|
||||
|
||||
* Synthesis of three complementary filters
|
||||
:PROPERTIES:
|
||||
:header-args:matlab+: :tangle matlab/4_three_complementary_filters.m
|
||||
:header-args:matlab+: :comments org :mkdirp yes
|
||||
:CUSTOM_ID: sec:three_comp_filters
|
||||
:END:
|
||||
|
||||
** Introduction :ignore:
|
||||
#+begin_note
|
||||
The Matlab file corresponding to this section is accessible [[file:matlab/4_three_complementary_filters.m][here]].
|
||||
#+end_note
|
||||
|
||||
In this section, the proposed synthesis method of complementary filters is generalized for the synthesis of a set of three complementary filters.
|
||||
|
||||
The complete Matlab script for this part is given in Section [[#sec:4_three_complementary_filters]].
|
||||
|
||||
** Matlab Init :noexport:ignore:
|
||||
#+begin_src matlab :tangle no :exports none :results silent :noweb yes :var current_dir=(file-name-directory buffer-file-name)
|
||||
@@ -1031,39 +1108,37 @@ freqs = logspace(-2, 3, 1000);
|
||||
|
||||
#+begin_src matlab :tangle no
|
||||
addpath('./matlab');
|
||||
addpath('./matlab/src');
|
||||
#+end_src
|
||||
|
||||
#+begin_src matlab :eval no
|
||||
addpath('./src');
|
||||
#+end_src
|
||||
** Synthesis Architecture
|
||||
The synthesis objective is to shape three filters that are complementary.
|
||||
This corresponds to the conditions eqref:eq:obj_three_cf where $W_1(s)$, $W_2(s)$ and $W_3(s)$ are weighting functions used to specify the maximum wanted magnitude of the three complementary filters.
|
||||
|
||||
** Theory
|
||||
We want:
|
||||
\begin{align*}
|
||||
& |H_1(j\omega)| < 1/|W_1(j\omega)|, \quad \forall\omega\\
|
||||
& |H_2(j\omega)| < 1/|W_2(j\omega)|, \quad \forall\omega\\
|
||||
& |H_3(j\omega)| < 1/|W_3(j\omega)|, \quad \forall\omega\\
|
||||
\begin{equation}
|
||||
\begin{aligned}
|
||||
& |H_1(j\omega)| < \frac{1}{|W_1(j\omega)|}, \quad \forall\omega \\
|
||||
& |H_2(j\omega)| < \frac{1}{|W_2(j\omega)|}, \quad \forall\omega \\
|
||||
& |H_3(j\omega)| < \frac{1}{|W_3(j\omega)|}, \quad \forall\omega \\
|
||||
& H_1(s) + H_2(s) + H_3(s) = 1
|
||||
\end{align*}
|
||||
\end{aligned} \label{eq:obj_three_cf}
|
||||
\end{equation}
|
||||
|
||||
For that, we use the $\mathcal{H}_\infty$ synthesis with the architecture shown on figure [[fig:comp_filter_three_hinf]].
|
||||
This synthesis can be done by performing the standard $\mathcal{H}_\infty$ synthesis with on the generalized plant in Figure [[fig:comp_filter_three_hinf]].
|
||||
|
||||
#+name: fig:comp_filter_three_hinf
|
||||
#+caption: Generalized architecture for generating 3 complementary filters
|
||||
[[file:figs-journal/comp_filter_three_hinf_fb.png]]
|
||||
|
||||
The $\mathcal{H}_\infty$ objective is:
|
||||
\begin{align*}
|
||||
& |(1 - H_2(j\omega) - H_3(j\omega)) W_1(j\omega)| < 1, \quad \forall\omega\\
|
||||
& |H_2(j\omega) W_2(j\omega)| < 1, \quad \forall\omega\\
|
||||
& |H_3(j\omega) W_3(j\omega)| < 1, \quad \forall\omega\\
|
||||
\end{align*}
|
||||
After synthesis, filter $H_2(s)$ and $H_3(s)$ are obtained as shown in Figure [[fig:comp_filter_three_hinf]].
|
||||
The last filter $H_1(s)$ is defined as the complementary of the two others as in eqref:eq:H1_complementary_of_H2_H3.
|
||||
|
||||
And thus if we choose $H_1 = 1 - H_2 - H_3$ we have solved the problem.
|
||||
\begin{equation}
|
||||
H_1(s) = 1 - H_2(s) - H_3(s) \label{eq:H1_complementary_of_H2_H3}
|
||||
\end{equation}
|
||||
|
||||
** Weights
|
||||
First we define the weights.
|
||||
The three weighting functions are defined as shown below.
|
||||
|
||||
#+begin_src matlab
|
||||
%% Design of the Weighting Functions
|
||||
W1 = generateWF('n', 2, 'w0', 2*pi*1, 'G0', 1/10, 'Ginf', 1000, 'Gc', 0.5);
|
||||
@@ -1071,6 +1146,8 @@ W2 = 0.22*(1 + s/2/pi/1)^2/(sqrt(1e-4) + s/2/pi/1)^2*(1 + s/2/pi/10)^2/(1 + s/2/
|
||||
W3 = generateWF('n', 3, 'w0', 2*pi*10, 'G0', 1000, 'Ginf', 1/10, 'Gc', 0.5);
|
||||
#+end_src
|
||||
|
||||
Their inverse magnitudes are displayed in Figure [[fig:three_weighting_functions]].
|
||||
|
||||
#+begin_src matlab :exports none
|
||||
%% Inverse magnitude of the weighting functions
|
||||
figure;
|
||||
@@ -1099,7 +1176,9 @@ exportFig('figs/three_weighting_functions.pdf', 'width', 'wide', 'height', 'norm
|
||||
[[file:figs/three_weighting_functions.png]]
|
||||
|
||||
** H-Infinity Synthesis
|
||||
Then we create the generalized plant =P=.
|
||||
|
||||
The generalized plant in Figure [[fig:comp_filter_three_hinf]] containing the weighting functions is defined below.
|
||||
|
||||
#+begin_src matlab
|
||||
%% Generalized plant for the synthesis of 3 complementary filters
|
||||
P = [W1 -W1 -W1;
|
||||
@@ -1108,7 +1187,7 @@ P = [W1 -W1 -W1;
|
||||
1 0 0];
|
||||
#+end_src
|
||||
|
||||
And we do the $\mathcal{H}_\infty$ synthesis.
|
||||
And the standard $\mathcal{H}_\infty$ synthesis using the =hinfsyn= command is performed.
|
||||
#+begin_src matlab :results output replace :exports both
|
||||
%% Standard H-Infinity Synthesis
|
||||
[H, ~, gamma, ~] = hinfsyn(P, 1, 2,'TOLGAM', 0.001, 'METHOD', 'ric', 'DISPLAY', 'on');
|
||||
@@ -1116,7 +1195,6 @@ And we do the $\mathcal{H}_\infty$ synthesis.
|
||||
|
||||
#+RESULTS:
|
||||
#+begin_example
|
||||
[H, ~, gamma, ~] = hinfsyn(P, 1, 2,'TOLGAM', 0.001, 'METHOD', 'ric', 'DISPLAY', 'on');
|
||||
Resetting value of Gamma min based on D_11, D_12, D_21 terms
|
||||
|
||||
Test bounds: 0.1000 < gamma <= 1050.0000
|
||||
@@ -1148,15 +1226,21 @@ Test bounds: 0.1000 < gamma <= 1050.0000
|
||||
Gamma value achieved: 0.9936
|
||||
#+end_example
|
||||
|
||||
** Obtained Complementary Filters
|
||||
The obtained filters are:
|
||||
#+begin_src matlab
|
||||
%%
|
||||
The two synthesized filters $H_2(s)$ and $H_3(s)$ are defined below:
|
||||
#+begin_src matlab :exports none
|
||||
%% Synthesized H2 and H3 filters
|
||||
H2 = tf(H(1));
|
||||
H3 = tf(H(2));
|
||||
#+end_src
|
||||
|
||||
And the third filter $H_1(s)$ is defined using eqref:eq:H1_complementary_of_H2_H3.
|
||||
#+begin_src matlab
|
||||
%% H1 is defined as the complementary filter of H2 and H3
|
||||
H1 = 1 - H2 - H3;
|
||||
#+end_src
|
||||
|
||||
The bode plots of the three obtained complementary filters are shown in Figure [[fig:three_complementary_filters_results]].
|
||||
|
||||
#+begin_src matlab :exports none
|
||||
%% Bode plot of the obtained complementary filters
|
||||
figure;
|
||||
@@ -1217,31 +1301,53 @@ exportFig('figs/three_complementary_filters_results.pdf', 'width', 'wide', 'heig
|
||||
#+RESULTS:
|
||||
[[file:figs/three_complementary_filters_results.png]]
|
||||
|
||||
* Bibliography :ignore:
|
||||
* Matlab Scripts
|
||||
:PROPERTIES:
|
||||
:CUSTOM_ID: sec:bibliography
|
||||
:CUSTOM_ID: sec:matlab_scripts
|
||||
:END:
|
||||
bibliographystyle:unsrt
|
||||
bibliography:ref.bib
|
||||
** =1_synthesis_complementary_filters.m=
|
||||
:PROPERTIES:
|
||||
:CUSTOM_ID: sec:1_synthesis_complementary_filters
|
||||
:END:
|
||||
This scripts corresponds to section 3 of cite:dehaeze21_new_method_desig_compl_filter.
|
||||
|
||||
* Functions
|
||||
#+INCLUDE: "matlab/1_synthesis_complementary_filters.m" src matlab -n
|
||||
|
||||
** =2_ligo_complementary_filters.m=
|
||||
:PROPERTIES:
|
||||
:header-args:matlab+: :comments none :mkdirp yes :eval no
|
||||
:CUSTOM_ID: sec:functions
|
||||
:CUSTOM_ID: sec:2_ligo_complementary_filters
|
||||
:END:
|
||||
|
||||
This scripts corresponds to section 4 of cite:dehaeze21_new_method_desig_compl_filter.
|
||||
|
||||
#+INCLUDE: "matlab/2_ligo_complementary_filters.m" src matlab -n
|
||||
|
||||
** =3_closed_loop_complementary_filters.m=
|
||||
:PROPERTIES:
|
||||
:CUSTOM_ID: sec:3_closed_loop_complementary_filters
|
||||
:END:
|
||||
|
||||
This scripts corresponds to section 5.1 of cite:dehaeze21_new_method_desig_compl_filter.
|
||||
|
||||
#+INCLUDE: "matlab/3_closed_loop_complementary_filters.m" src matlab -n
|
||||
|
||||
** =4_three_complementary_filters.m=
|
||||
:PROPERTIES:
|
||||
:CUSTOM_ID: sec:4_three_complementary_filters
|
||||
:END:
|
||||
|
||||
This scripts corresponds to section 5.2 of cite:dehaeze21_new_method_desig_compl_filter.
|
||||
|
||||
#+INCLUDE: "matlab/4_three_complementary_filters.m" src matlab -n
|
||||
|
||||
** =generateWF=: Generate Weighting Functions
|
||||
:PROPERTIES:
|
||||
:header-args:matlab+: :tangle matlab/src/generateWF.m
|
||||
:header-args:matlab+: :tangle matlab/generateWF.m
|
||||
:header-args:matlab+: :comments none :mkdirp yes :eval no
|
||||
:CUSTOM_ID: sec:generateWF
|
||||
:END:
|
||||
<<sec:generateWF>>
|
||||
|
||||
This Matlab function is accessible [[file:matlab/src/generateWF.m][here]].
|
||||
|
||||
*** Function description
|
||||
:PROPERTIES:
|
||||
:UNNUMBERED: t
|
||||
:END:
|
||||
This function is used to easily generate weighting functions from classical requirements.
|
||||
|
||||
#+begin_src matlab
|
||||
function [W] = generateWF(args)
|
||||
@@ -1258,13 +1364,7 @@ function [W] = generateWF(args)
|
||||
%
|
||||
% Outputs:
|
||||
% - W - Generated Weighting Function
|
||||
#+end_src
|
||||
|
||||
*** Optional Parameters
|
||||
:PROPERTIES:
|
||||
:UNNUMBERED: t
|
||||
:END:
|
||||
#+begin_src matlab
|
||||
%% Argument validation
|
||||
arguments
|
||||
args.n (1,1) double {mustBeInteger, mustBePositive} = 1
|
||||
@@ -1273,54 +1373,19 @@ arguments
|
||||
args.Gc (1,1) double {mustBeNumeric, mustBePositive} = 1
|
||||
args.w0 (1,1) double {mustBeNumeric, mustBePositive} = 1
|
||||
end
|
||||
#+end_src
|
||||
|
||||
Verification that the parameters $G_0$, $G_c$ and $G_\infty$ are satisfy condition eqref:eq:cond_formula_1 or eqref:eq:cond_formula_2.
|
||||
\begin{equation}
|
||||
G_0 < 1 < G_\infty \text{ and } G_0 < G_c < G_\infty \label{eq:cond_formula_1}
|
||||
\end{equation}
|
||||
\begin{equation}
|
||||
G_\infty < 1 < G_0 \text{ and } G_\infty < G_c < G_0 \label{eq:cond_formula_2}
|
||||
\end{equation}
|
||||
|
||||
#+begin_src matlab
|
||||
% Verification of correct relation between G0, Gc and Ginf
|
||||
mustBeBetween(args.G0, args.Gc, args.Ginf);
|
||||
#+end_src
|
||||
|
||||
*** Generate the Weighting function
|
||||
:PROPERTIES:
|
||||
:UNNUMBERED: t
|
||||
:END:
|
||||
#+begin_src matlab
|
||||
%% Initialize the Laplace variable
|
||||
s = zpk('s');
|
||||
#+end_src
|
||||
|
||||
The weighting function formula use is:
|
||||
#+name: eq:weight_formula
|
||||
\begin{equation}
|
||||
W(s) = \left( \frac{
|
||||
\frac{1}{\omega_c} \sqrt{\frac{1 - \left(\frac{G_0}{G_c}\right)^{\frac{2}{n}}}{1 - \left(\frac{G_c}{G_\infty}\right)^{\frac{2}{n}}}} s + \left(\frac{G_0}{G_c}\right)^{\frac{1}{n}}
|
||||
}{
|
||||
\left(\frac{1}{G_\infty}\right)^{\frac{1}{n}} \frac{1}{\omega_c} \sqrt{\frac{1 - \left(\frac{G_0}{G_c}\right)^{\frac{2}{n}}}{1 - \left(\frac{G_c}{G_\infty}\right)^{\frac{2}{n}}}} s + \left(\frac{1}{G_c}\right)^{\frac{1}{n}}
|
||||
}\right)^n
|
||||
\end{equation}
|
||||
|
||||
#+begin_src matlab
|
||||
%% Create the weighting function according to formula
|
||||
W = (((1/args.w0)*sqrt((1-(args.G0/args.Gc)^(2/args.n))/(1-(args.Gc/args.Ginf)^(2/args.n)))*s + ...
|
||||
(args.G0/args.Gc)^(1/args.n))/...
|
||||
((1/args.Ginf)^(1/args.n)*(1/args.w0)*sqrt((1-(args.G0/args.Gc)^(2/args.n))/(1-(args.Gc/args.Ginf)^(2/args.n)))*s + ...
|
||||
(1/args.Gc)^(1/args.n)))^args.n;
|
||||
#+end_src
|
||||
|
||||
|
||||
*** Verification of the $G_0$, $G_c$ and $G_\infty$ gains
|
||||
:PROPERTIES:
|
||||
:UNNUMBERED: t
|
||||
:END:
|
||||
#+begin_src matlab
|
||||
%% Custom validation function
|
||||
function mustBeBetween(a,b,c)
|
||||
if ~((a > b && b > c) || (c > b && b > a))
|
||||
@@ -1332,16 +1397,12 @@ function mustBeBetween(a,b,c)
|
||||
|
||||
** =generateCF=: Generate Complementary Filters
|
||||
:PROPERTIES:
|
||||
:header-args:matlab+: :tangle matlab/src/generateCF.m
|
||||
:header-args:matlab+: :tangle matlab/generateCF.m
|
||||
:header-args:matlab+: :comments none :mkdirp yes :eval no
|
||||
:CUSTOM_ID: sec:generateCF
|
||||
:END:
|
||||
<<sec:generateCF>>
|
||||
|
||||
This Matlab function is accessible [[file:matlab/src/generateCF.m][here]].
|
||||
|
||||
*** Function description
|
||||
:PROPERTIES:
|
||||
:UNNUMBERED: t
|
||||
:END:
|
||||
This function is used to easily synthesize a set of two complementary filters using the $\mathcal{H}_\infty$ synthesis.
|
||||
|
||||
#+begin_src matlab
|
||||
function [H1, H2] = generateCF(W1, W2, args)
|
||||
@@ -1359,13 +1420,7 @@ function [H1, H2] = generateCF(W1, W2, args)
|
||||
% Outputs:
|
||||
% - H1 - Generated H1 Filter
|
||||
% - H2 - Generated H2 Filter
|
||||
#+end_src
|
||||
|
||||
*** Optional Parameters
|
||||
:PROPERTIES:
|
||||
:UNNUMBERED: t
|
||||
:END:
|
||||
#+begin_src matlab
|
||||
%% Argument validation
|
||||
arguments
|
||||
W1
|
||||
@@ -1373,25 +1428,21 @@ arguments
|
||||
args.method char {mustBeMember(args.method,{'lmi', 'ric'})} = 'ric'
|
||||
args.display char {mustBeMember(args.display,{'on', 'off'})} = 'on'
|
||||
end
|
||||
#+end_src
|
||||
|
||||
*** H-Infinity Synthesis
|
||||
:PROPERTIES:
|
||||
:UNNUMBERED: t
|
||||
:END:
|
||||
#+begin_src matlab
|
||||
%% The generalized plant is defined
|
||||
P = [W1 -W1;
|
||||
0 W2;
|
||||
1 0];
|
||||
#+end_src
|
||||
|
||||
#+begin_src matlab :results output replace :exports both
|
||||
%% The standard H-infinity synthesis is performed
|
||||
[H2, ~, gamma, ~] = hinfsyn(P, 1, 1,'TOLGAM', 0.001, 'METHOD', args.method, 'DISPLAY', args.display);
|
||||
#+end_src
|
||||
|
||||
#+begin_src matlab
|
||||
%% H1 is defined as the complementary of H2
|
||||
H1 = 1 - H2;
|
||||
#+end_src
|
||||
|
||||
* Bibliography :ignore:
|
||||
# For html export only:
|
||||
# bibliography:ref.bib
|
||||
|
||||
#+latex: \printbibliography
|
||||
|
||||
BIN
matlab/dehaeze21_desig_compl_filte_matlab.pdf
Normal file
BIN
matlab/dehaeze21_desig_compl_filte_matlab.pdf
Normal file
Binary file not shown.
1404
matlab/dehaeze21_desig_compl_filte_matlab.tex
Normal file
1404
matlab/dehaeze21_desig_compl_filte_matlab.tex
Normal file
File diff suppressed because it is too large
Load Diff
BIN
matlab/index.pdf
BIN
matlab/index.pdf
Binary file not shown.
@@ -7,9 +7,6 @@ s = zpk('s');
|
||||
%% Initialize Frequency Vector
|
||||
freqs = logspace(-1, 3, 1000);
|
||||
|
||||
%% Add functions to path
|
||||
addpath('./src');
|
||||
|
||||
%% Weighting Function Design
|
||||
% Parameters
|
||||
n = 3; w0 = 2*pi*10; G0 = 1e-3; G1 = 1e1; Gc = 2;
|
||||
@@ -45,8 +42,8 @@ ylim([5e-4, 20]);
|
||||
yticks([1e-4, 1e-3, 1e-2, 1e-1, 1, 1e1]);
|
||||
|
||||
%% Design of the Weighting Functions
|
||||
W1 = generateWF('n', 3, 'w0', 2*pi*10, 'G0', 1000, 'G1', 1/10, 'Gc', 0.45);
|
||||
W2 = generateWF('n', 2, 'w0', 2*pi*10, 'G0', 1/10, 'G1', 1000, 'Gc', 0.45);
|
||||
W1 = generateWF('n', 3, 'w0', 2*pi*10, 'G0', 1000, 'Ginf', 1/10, 'Gc', 0.45);
|
||||
W2 = generateWF('n', 2, 'w0', 2*pi*10, 'G0', 1/10, 'Ginf', 1000, 'Gc', 0.45);
|
||||
|
||||
%% Plot of the Weighting function magnitude
|
||||
figure;
|
||||
|
||||
@@ -7,9 +7,6 @@ s = zpk('s');
|
||||
%% Initialize Frequency Vector
|
||||
freqs = logspace(-3, 0, 1000);
|
||||
|
||||
%% Add functions to path
|
||||
addpath('./src');
|
||||
|
||||
%% Upper bounds for the complementary filters
|
||||
figure;
|
||||
hold on;
|
||||
|
||||
@@ -7,12 +7,9 @@ s = zpk('s');
|
||||
%% Initialize Frequency Vector
|
||||
freqs = logspace(-1, 3, 1000);
|
||||
|
||||
%% Add functions to path
|
||||
addpath('./src');
|
||||
|
||||
%% Design of the Weighting Functions
|
||||
W1 = generateWF('n', 3, 'w0', 2*pi*10, 'G0', 1000, 'G1', 1/10, 'Gc', 0.45);
|
||||
W2 = generateWF('n', 2, 'w0', 2*pi*10, 'G0', 1/10, 'G1', 1000, 'Gc', 0.45);
|
||||
W1 = generateWF('n', 3, 'w0', 2*pi*10, 'G0', 1000, 'Ginf', 1/10, 'Gc', 0.45);
|
||||
W2 = generateWF('n', 2, 'w0', 2*pi*10, 'G0', 1/10, 'Ginf', 1000, 'Gc', 0.45);
|
||||
|
||||
%% Generalized plant for "closed-loop" complementary filter synthesis
|
||||
P = [ W1 0 1;
|
||||
|
||||
@@ -6,15 +6,10 @@ s = zpk('s');
|
||||
|
||||
freqs = logspace(-2, 3, 1000);
|
||||
|
||||
addpath('./src');
|
||||
|
||||
% Weights
|
||||
% First we define the weights.
|
||||
|
||||
%% Design of the Weighting Functions
|
||||
W1 = generateWF('n', 2, 'w0', 2*pi*1, 'G0', 1/10, 'G1', 1000, 'Gc', 0.5);
|
||||
W1 = generateWF('n', 2, 'w0', 2*pi*1, 'G0', 1/10, 'Ginf', 1000, 'Gc', 0.5);
|
||||
W2 = 0.22*(1 + s/2/pi/1)^2/(sqrt(1e-4) + s/2/pi/1)^2*(1 + s/2/pi/10)^2/(1 + s/2/pi/1000)^2;
|
||||
W3 = generateWF('n', 3, 'w0', 2*pi*10, 'G0', 1000, 'G1', 1/10, 'Gc', 0.5);
|
||||
W3 = generateWF('n', 3, 'w0', 2*pi*10, 'G0', 1000, 'Ginf', 1/10, 'Gc', 0.5);
|
||||
|
||||
%% Inverse magnitude of the weighting functions
|
||||
figure;
|
||||
@@ -32,28 +27,20 @@ xlim([freqs(1), freqs(end)]); ylim([2e-4, 1.3e1])
|
||||
leg = legend('location', 'northeast', 'FontSize', 8);
|
||||
leg.ItemTokenSize(1) = 18;
|
||||
|
||||
% H-Infinity Synthesis
|
||||
% Then we create the generalized plant =P=.
|
||||
|
||||
%% Generalized plant for the synthesis of 3 complementary filters
|
||||
P = [W1 -W1 -W1;
|
||||
0 W2 0 ;
|
||||
0 0 W3;
|
||||
1 0 0];
|
||||
|
||||
|
||||
|
||||
% And we do the $\mathcal{H}_\infty$ synthesis.
|
||||
|
||||
%% Standard H-Infinity Synthesis
|
||||
[H, ~, gamma, ~] = hinfsyn(P, 1, 2,'TOLGAM', 0.001, 'METHOD', 'ric', 'DISPLAY', 'on');
|
||||
|
||||
% Obtained Complementary Filters
|
||||
% The obtained filters are:
|
||||
|
||||
%%
|
||||
%% Synthesized H2 and H3 filters
|
||||
H2 = tf(H(1));
|
||||
H3 = tf(H(2));
|
||||
|
||||
%% H1 is defined as the complementary filter of H2 and H3
|
||||
H1 = 1 - H2 - H3;
|
||||
|
||||
%% Bode plot of the obtained complementary filters
|
||||
|
||||
114
matlab/preamble.tex
Normal file
114
matlab/preamble.tex
Normal file
@@ -0,0 +1,114 @@
|
||||
\usepackage{float}
|
||||
|
||||
\usepackage{caption,tabularx,booktabs}
|
||||
|
||||
\usepackage{biblatex}
|
||||
|
||||
\usepackage{fontawesome}
|
||||
|
||||
\usepackage{caption}
|
||||
\usepackage{subcaption}
|
||||
|
||||
\captionsetup[figure]{labelfont=bf}
|
||||
\captionsetup[subfigure]{labelfont=bf}
|
||||
\captionsetup[listing]{labelfont=bf}
|
||||
\captionsetup[table]{labelfont=bf}
|
||||
|
||||
\usepackage{xcolor}
|
||||
|
||||
\definecolor{my-blue}{HTML}{6b7adb}
|
||||
\definecolor{my-pale-blue}{HTML}{e6e9f9}
|
||||
\definecolor{my-red}{HTML}{db6b6b}
|
||||
\definecolor{my-pale-red}{HTML}{f9e6e6}
|
||||
\definecolor{my-green}{HTML}{6bdbb6}
|
||||
\definecolor{my-pale-green}{HTML}{e6f9f3}
|
||||
\definecolor{my-yellow}{HTML}{dbd26b}
|
||||
\definecolor{my-pale-yellow}{HTML}{f9f7e6}
|
||||
\definecolor{my-orange}{HTML}{dba76b}
|
||||
\definecolor{my-pale-orange}{HTML}{f9f0e6}
|
||||
\definecolor{my-grey}{HTML}{a3a3a3}
|
||||
\definecolor{my-pale-grey}{HTML}{f0f0f0}
|
||||
\definecolor{my-turq}{HTML}{6bc7db}
|
||||
\definecolor{my-pale-turq}{HTML}{e6f6f9}
|
||||
|
||||
\usepackage{inconsolata}
|
||||
|
||||
\usepackage[newfloat=true, chapter]{minted}
|
||||
\usemintedstyle{autumn}
|
||||
|
||||
\setminted{frame=lines,breaklines=true,tabsize=4,fontsize=\scriptsize,autogobble=true,labelposition=topline,bgcolor=my-pale-grey}
|
||||
\setminted[matlab]{label=Matlab}
|
||||
\setminted[latex]{label=LaTeX}
|
||||
\setminted[bash]{label=Bash}
|
||||
\setminted[python]{label=Python}
|
||||
\setminted[text]{label=Results}
|
||||
\setminted[md]{label=Org Mode}
|
||||
|
||||
\setmintedinline{fontsize=\normalsize,bgcolor=my-pale-grey}
|
||||
|
||||
\usepackage[most]{tcolorbox}
|
||||
|
||||
\tcbuselibrary{minted}
|
||||
|
||||
\newtcolorbox{seealso}{ enhanced,breakable,colback=my-pale-grey,colframe=my-grey,fonttitle=\bfseries,title=See Also}
|
||||
\newtcolorbox{hint}{ enhanced,breakable,colback=my-pale-grey,colframe=my-grey,fonttitle=\bfseries,title=Hint}
|
||||
\newtcolorbox{definition}{enhanced,breakable,colback=my-pale-red, colframe=my-red, fonttitle=\bfseries,title=Definition}
|
||||
\newtcolorbox{important}{ enhanced,breakable,colback=my-pale-red, colframe=my-red, fonttitle=\bfseries,title=Important}
|
||||
\newtcolorbox{exampl}[1][]{ enhanced,breakable,colback=my-pale-green,colframe=my-green,fonttitle=\bfseries,title=Example,#1}
|
||||
\newtcolorbox{exercice}{ enhanced,breakable,colback=my-pale-yellow,colframe=my-yellow,fonttitle=\bfseries,title=Exercice}
|
||||
\newtcolorbox{question}{ enhanced,breakable,colback=my-pale-yellow,colframe=my-yellow,fonttitle=\bfseries,title=Question}
|
||||
\newtcolorbox{answer}{ enhanced,breakable,colback=my-pale-turq,colframe=my-turq,fonttitle=\bfseries,title=Answer}
|
||||
\newtcolorbox{summary}{ enhanced,breakable,colback=my-pale-blue,colframe=my-blue,fonttitle=\bfseries,title=Summary}
|
||||
\newtcolorbox{note}{ enhanced,breakable,colback=my-pale-blue,colframe=my-blue,fonttitle=\bfseries,title=Note}
|
||||
\newtcolorbox{caution}{ enhanced,breakable,colback=my-pale-orange,colframe=my-orange,fonttitle=\bfseries,title=Caution}
|
||||
\newtcolorbox{warning}{ enhanced,breakable,colback=my-pale-orange,colframe=my-orange,fonttitle=\bfseries,title=Warning}
|
||||
|
||||
\newtcolorbox{my-quote}[1]{%
|
||||
colback=my-pale-grey,
|
||||
grow to right by=-10mm,
|
||||
grow to left by=-10mm,
|
||||
boxrule=0pt,
|
||||
boxsep=0pt,
|
||||
breakable,
|
||||
enhanced jigsaw,
|
||||
borderline west={4pt}{0pt}{my-grey}}
|
||||
|
||||
\renewenvironment{quote}{\begin{my-quote}}{\end{my-quote}}
|
||||
|
||||
\newtcolorbox{my-verse}[1]{%
|
||||
colback=my-pale-grey,
|
||||
grow to right by=-10mm,
|
||||
grow to left by=-10mm,
|
||||
boxrule=0pt,
|
||||
boxsep=0pt,
|
||||
breakable,
|
||||
enhanced jigsaw,
|
||||
borderline west={4pt}{0pt}{my-grey}}
|
||||
|
||||
\renewenvironment{verse}{\begin{my-verse}}{\end{my-verse}}
|
||||
|
||||
\usepackage{environ}% http://ctan.org/pkg/environ
|
||||
\NewEnviron{aside}{%
|
||||
\marginpar{\BODY}
|
||||
}
|
||||
|
||||
\renewenvironment{verbatim}{\VerbatimEnvironment\begin{minted}[]{text}}{\end{minted}}
|
||||
|
||||
\usepackage{soul}
|
||||
\sethlcolor{my-pale-grey}
|
||||
|
||||
\let\OldTexttt\texttt
|
||||
\renewcommand{\texttt}[1]{{\ttfamily\hl{\mbox{\,#1\,}}}}
|
||||
|
||||
\makeatletter
|
||||
\preto\Gin@extensions{png,}
|
||||
\DeclareGraphicsRule{.png}{pdf}{.pdf}{\noexpand\Gin@base.pdf}
|
||||
\makeatother
|
||||
|
||||
\usepackage{hyperref}
|
||||
\hypersetup{
|
||||
colorlinks = true,
|
||||
allcolors = my-blue
|
||||
}
|
||||
|
||||
\usepackage{hypcap}
|
||||
@@ -1,16 +1,63 @@
|
||||
@phdthesis{hua05_low_ligo,
|
||||
author = {Hua, Wensheng},
|
||||
school = {stanford university},
|
||||
title = {Low frequency vibration isolation and alignment system for
|
||||
advanced LIGO},
|
||||
year = 2005,
|
||||
@inproceedings{hua04_low_ligo,
|
||||
author = {Hua, Wensheng and Adhikari, R and DeBra, Daniel B and
|
||||
Giaime, Joseph A and Hammond, Giles Dominic and Hardham, C and
|
||||
Hennessy, Mike and How, Jonathan P and Lantz, Brian T and
|
||||
Macinnis, M and others},
|
||||
title = {Low-frequency active vibration isolation for advanced
|
||||
{LIGO}},
|
||||
booktitle = {Gravitational Wave and Particle Astrophysics Detectors},
|
||||
year = 2004,
|
||||
volume = 5500,
|
||||
pages = {194--205},
|
||||
organization = {International Society for Optics and Photonics},
|
||||
}
|
||||
|
||||
@article{dehaeze21_new_method_desig_compl_filter,
|
||||
author = {Dehaeze, Thomas and Vermat, Mohit and Collette, Christophe},
|
||||
title = {A New Method of Designing Complementary Filters for Sensor
|
||||
Fusion Using the $\mathcal{H}_\infty$ Synthesis},
|
||||
Fusion Using the H-Infinity Synthesis},
|
||||
journal = {Mechanical Systems and Signal Processing},
|
||||
year = 2021,
|
||||
month = {Nov},
|
||||
}
|
||||
|
||||
@book{matlab20,
|
||||
author = {MATLAB},
|
||||
title = {version 9.9.0 (R2020b)},
|
||||
year = 2020,
|
||||
publisher = {The MathWorks Inc.},
|
||||
address = {Natick, Massachusetts},
|
||||
}
|
||||
|
||||
@article{sturm99_using_sedum,
|
||||
author = {Sturm, Jos F},
|
||||
title = {Using {SeDuMi} 1.02, a Matlab Toolbox for Optimization Over
|
||||
Symmetric Cones},
|
||||
journal = {Optimization methods and software},
|
||||
volume = 11,
|
||||
number = {1-4},
|
||||
pages = {625--653},
|
||||
year = 1999,
|
||||
doi = {10.1080/10556789908805766},
|
||||
url = {https://doi.org/10.1080/10556789908805766},
|
||||
publisher = {Taylor \& Francis},
|
||||
}
|
||||
|
||||
@inproceedings{hua04_polyp_fir_compl_filter_contr_system,
|
||||
author = {Hua, Wensheng and Debra, B. and Hardham, T. and Lantz, T.
|
||||
and Giaime, A.},
|
||||
title = {Polyphase {FIR} Complementary Filters for Control Systems},
|
||||
booktitle = {Proceedings of ASPE Spring Topical Meeting on Control of
|
||||
Precision Systems},
|
||||
year = 2004,
|
||||
pages = {109--114},
|
||||
}
|
||||
|
||||
@misc{grant14_cvx,
|
||||
author = {Michael Grant and Stephen Boyd},
|
||||
howpublished = {\url{http://cvxr.com/cvx}},
|
||||
month = 3,
|
||||
title = {{CVX}: Matlab Software for Disciplined Convex Programming,
|
||||
version 2.1},
|
||||
year = 2014,
|
||||
}
|
||||
|
||||
File diff suppressed because it is too large
Load Diff
1056
mohit/elsarticle.cls
1056
mohit/elsarticle.cls
File diff suppressed because it is too large
Load Diff
Binary file not shown.
Binary file not shown.
Binary file not shown.
Binary file not shown.
Binary file not shown.
File diff suppressed because it is too large
Load Diff
@@ -1,951 +0,0 @@
|
||||
%PDF-1.4
|
||||
%<25><><EFBFBD><EFBFBD>
|
||||
1 0 obj
|
||||
<<
|
||||
/Producer (Apache FOP Version 2.4.0-SNAPSHOT: PDFDocumentGraphics2D)
|
||||
/CreationDate (D:20201116104636+01'00')
|
||||
>>
|
||||
endobj
|
||||
2 0 obj
|
||||
<<
|
||||
/Type /ExtGState
|
||||
/CA 0.2509804
|
||||
>>
|
||||
endobj
|
||||
3 0 obj
|
||||
<<
|
||||
/Type /ExtGState
|
||||
/CA 0.2509804
|
||||
>>
|
||||
endobj
|
||||
4 0 obj
|
||||
<<
|
||||
/Type /ExtGState
|
||||
/CA 0.2509804
|
||||
>>
|
||||
endobj
|
||||
5 0 obj
|
||||
<<
|
||||
/Type /ExtGState
|
||||
/CA 0.2509804
|
||||
>>
|
||||
endobj
|
||||
6 0 obj
|
||||
<<
|
||||
/Type /ExtGState
|
||||
/CA 0.2509804
|
||||
>>
|
||||
endobj
|
||||
7 0 obj
|
||||
<<
|
||||
/Type /ExtGState
|
||||
/CA 0.2509804
|
||||
>>
|
||||
endobj
|
||||
8 0 obj
|
||||
<<
|
||||
/Type /ExtGState
|
||||
/CA 0.2509804
|
||||
>>
|
||||
endobj
|
||||
9 0 obj
|
||||
<<
|
||||
/Type /ExtGState
|
||||
/CA 0.2509804
|
||||
>>
|
||||
endobj
|
||||
10 0 obj
|
||||
<<
|
||||
/Type /ExtGState
|
||||
/CA 0.2509804
|
||||
>>
|
||||
endobj
|
||||
11 0 obj
|
||||
<<
|
||||
/Type /ExtGState
|
||||
/CA 0.2509804
|
||||
>>
|
||||
endobj
|
||||
12 0 obj
|
||||
<<
|
||||
/Type /ExtGState
|
||||
/CA 0.2509804
|
||||
>>
|
||||
endobj
|
||||
13 0 obj
|
||||
<<
|
||||
/Type /ExtGState
|
||||
/CA 0.2509804
|
||||
>>
|
||||
endobj
|
||||
14 0 obj
|
||||
<<
|
||||
/Type /ExtGState
|
||||
/CA 0.2509804
|
||||
>>
|
||||
endobj
|
||||
15 0 obj
|
||||
<<
|
||||
/Type /ExtGState
|
||||
/CA 0.2509804
|
||||
>>
|
||||
endobj
|
||||
16 0 obj
|
||||
<<
|
||||
/Type /ExtGState
|
||||
/CA 0.2509804
|
||||
>>
|
||||
endobj
|
||||
17 0 obj
|
||||
<<
|
||||
/Type /ExtGState
|
||||
/CA 0.2509804
|
||||
>>
|
||||
endobj
|
||||
18 0 obj
|
||||
<<
|
||||
/Type /ExtGState
|
||||
/CA 0.2509804
|
||||
>>
|
||||
endobj
|
||||
19 0 obj
|
||||
<<
|
||||
/Type /ExtGState
|
||||
/CA 0.2509804
|
||||
>>
|
||||
endobj
|
||||
20 0 obj
|
||||
<<
|
||||
/Type /ExtGState
|
||||
/CA 0.2509804
|
||||
>>
|
||||
endobj
|
||||
21 0 obj
|
||||
<<
|
||||
/Type /ExtGState
|
||||
/CA 0.2509804
|
||||
>>
|
||||
endobj
|
||||
22 0 obj
|
||||
<<
|
||||
/Type /ExtGState
|
||||
/CA 0.2509804
|
||||
>>
|
||||
endobj
|
||||
23 0 obj
|
||||
<<
|
||||
/Type /ExtGState
|
||||
/CA 0.2509804
|
||||
>>
|
||||
endobj
|
||||
24 0 obj
|
||||
<<
|
||||
/Type /ExtGState
|
||||
/CA 0.2509804
|
||||
>>
|
||||
endobj
|
||||
25 0 obj
|
||||
<<
|
||||
/Type /ExtGState
|
||||
/CA 0.2509804
|
||||
>>
|
||||
endobj
|
||||
26 0 obj
|
||||
<<
|
||||
/Type /ExtGState
|
||||
/CA 0.14901961
|
||||
>>
|
||||
endobj
|
||||
27 0 obj
|
||||
<<
|
||||
/Type /ExtGState
|
||||
/CA 0.14901961
|
||||
>>
|
||||
endobj
|
||||
28 0 obj
|
||||
<<
|
||||
/Type /ExtGState
|
||||
/CA 0.14901961
|
||||
>>
|
||||
endobj
|
||||
29 0 obj
|
||||
<<
|
||||
/Type /ExtGState
|
||||
/CA 0.14901961
|
||||
>>
|
||||
endobj
|
||||
30 0 obj
|
||||
<<
|
||||
/Type /ExtGState
|
||||
/CA 0.2509804
|
||||
>>
|
||||
endobj
|
||||
31 0 obj
|
||||
<<
|
||||
/Type /ExtGState
|
||||
/CA 0.2509804
|
||||
>>
|
||||
endobj
|
||||
32 0 obj
|
||||
<<
|
||||
/Type /ExtGState
|
||||
/CA 0.2509804
|
||||
>>
|
||||
endobj
|
||||
33 0 obj
|
||||
<<
|
||||
/Type /ExtGState
|
||||
/CA 0.2509804
|
||||
>>
|
||||
endobj
|
||||
34 0 obj
|
||||
<<
|
||||
/Type /ExtGState
|
||||
/CA 0.2509804
|
||||
>>
|
||||
endobj
|
||||
35 0 obj
|
||||
<<
|
||||
/Type /ExtGState
|
||||
/CA 0.2509804
|
||||
>>
|
||||
endobj
|
||||
36 0 obj
|
||||
<<
|
||||
/Type /ExtGState
|
||||
/CA 0.2509804
|
||||
>>
|
||||
endobj
|
||||
37 0 obj
|
||||
<<
|
||||
/Type /ExtGState
|
||||
/CA 0.2509804
|
||||
>>
|
||||
endobj
|
||||
38 0 obj
|
||||
<<
|
||||
/Type /ExtGState
|
||||
/CA 0.2509804
|
||||
>>
|
||||
endobj
|
||||
39 0 obj
|
||||
<<
|
||||
/Type /ExtGState
|
||||
/CA 0.2509804
|
||||
>>
|
||||
endobj
|
||||
40 0 obj
|
||||
<<
|
||||
/Type /ExtGState
|
||||
/CA 0.2509804
|
||||
>>
|
||||
endobj
|
||||
41 0 obj
|
||||
<<
|
||||
/Type /ExtGState
|
||||
/CA 0.2509804
|
||||
>>
|
||||
endobj
|
||||
42 0 obj
|
||||
<<
|
||||
/Type /ExtGState
|
||||
/CA 0.2509804
|
||||
>>
|
||||
endobj
|
||||
43 0 obj
|
||||
<<
|
||||
/Type /ExtGState
|
||||
/CA 0.2509804
|
||||
>>
|
||||
endobj
|
||||
44 0 obj
|
||||
<<
|
||||
/Type /ExtGState
|
||||
/CA 0.2509804
|
||||
>>
|
||||
endobj
|
||||
45 0 obj
|
||||
<<
|
||||
/Type /ExtGState
|
||||
/CA 0.2509804
|
||||
>>
|
||||
endobj
|
||||
46 0 obj
|
||||
<<
|
||||
/Type /ExtGState
|
||||
/CA 0.2509804
|
||||
>>
|
||||
endobj
|
||||
47 0 obj
|
||||
<<
|
||||
/Type /ExtGState
|
||||
/CA 0.2509804
|
||||
>>
|
||||
endobj
|
||||
48 0 obj
|
||||
<<
|
||||
/Type /ExtGState
|
||||
/CA 0.2509804
|
||||
>>
|
||||
endobj
|
||||
49 0 obj
|
||||
<<
|
||||
/Type /ExtGState
|
||||
/CA 0.2509804
|
||||
>>
|
||||
endobj
|
||||
50 0 obj
|
||||
<<
|
||||
/Type /ExtGState
|
||||
/CA 0.2509804
|
||||
>>
|
||||
endobj
|
||||
51 0 obj
|
||||
<<
|
||||
/Type /ExtGState
|
||||
/CA 0.2509804
|
||||
>>
|
||||
endobj
|
||||
52 0 obj
|
||||
<<
|
||||
/Type /ExtGState
|
||||
/CA 0.2509804
|
||||
>>
|
||||
endobj
|
||||
53 0 obj
|
||||
<<
|
||||
/Type /ExtGState
|
||||
/CA 0.2509804
|
||||
>>
|
||||
endobj
|
||||
54 0 obj
|
||||
<<
|
||||
/Type /ExtGState
|
||||
/CA 0.2509804
|
||||
>>
|
||||
endobj
|
||||
55 0 obj
|
||||
<<
|
||||
/Type /ExtGState
|
||||
/CA 0.2509804
|
||||
>>
|
||||
endobj
|
||||
56 0 obj
|
||||
<<
|
||||
/Type /ExtGState
|
||||
/CA 0.2509804
|
||||
>>
|
||||
endobj
|
||||
57 0 obj
|
||||
<<
|
||||
/Type /ExtGState
|
||||
/CA 0.2509804
|
||||
>>
|
||||
endobj
|
||||
58 0 obj
|
||||
<<
|
||||
/Type /ExtGState
|
||||
/CA 0.2509804
|
||||
>>
|
||||
endobj
|
||||
59 0 obj
|
||||
<<
|
||||
/Type /ExtGState
|
||||
/CA 0.2509804
|
||||
>>
|
||||
endobj
|
||||
60 0 obj
|
||||
<<
|
||||
/Type /ExtGState
|
||||
/CA 0.2509804
|
||||
>>
|
||||
endobj
|
||||
61 0 obj
|
||||
<<
|
||||
/Type /ExtGState
|
||||
/CA 0.2509804
|
||||
>>
|
||||
endobj
|
||||
62 0 obj
|
||||
<<
|
||||
/Type /ExtGState
|
||||
/CA 0.14901961
|
||||
>>
|
||||
endobj
|
||||
63 0 obj
|
||||
<<
|
||||
/Type /ExtGState
|
||||
/CA 0.14901961
|
||||
>>
|
||||
endobj
|
||||
64 0 obj
|
||||
<<
|
||||
/Type /ExtGState
|
||||
/CA 0.14901961
|
||||
>>
|
||||
endobj
|
||||
65 0 obj
|
||||
<<
|
||||
/Type /ExtGState
|
||||
/CA 0.14901961
|
||||
>>
|
||||
endobj
|
||||
66 0 obj
|
||||
<<
|
||||
/Type /ExtGState
|
||||
/CA 0.14901961
|
||||
>>
|
||||
endobj
|
||||
67 0 obj
|
||||
<< /Length 68 0 R /Filter /FlateDecode >>
|
||||
stream
|
||||
x<EFBFBD>Žˎ$;<3B> 8?_<>?P.#<23>O@<10><>- <10><><EFBFBD>k&hԍ+<2B>5zп<7A><D0BF><EFBFBD>̓<EFBFBD><CD83>:Y<><59>Y<>p+<2B>WҝF#7<>{<7B><>?<3F><><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>_<5F>㸧\<5C><><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>#<23><>(<28><>Q<EFBFBD>Y<EFBFBD><59> <1D><><EFBFBD>R<EFBFBD>e<EFBFBD><65>zϷ<7A><CFB7><EFBFBD>>ZO<5A><4F>Ɓ<EFBFBD><C681>?<3F>8n<38>鏔<EFBFBD><E98F94><EFBFBD>y<EFBFBD><79> #U<><55>M:;<3B>@P<50><7F><EFBFBD><7F> <1F><><EFBFBD><EFBFBD><7F><EFBFBD><EFBFBD><EFBFBD><13><><EFBFBD><EFBFBD>!<21><> =<3D><><EFBFBD>-<2D><><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>G<EFBFBD><47>?<><7F>Y<EFBFBD><59><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD> i<><69>q<EFBFBD>V<EFBFBD>y<EFBFBD><14><>4<EFBFBD><34><EFBFBD><EFBFBD><EFBFBD>V<EFBFBD><56><EFBFBD>r <1E><03><><18><><EFBFBD><EFBFBD><EFBFBD>K<EFBFBD>u~}ZK<>g <1E>]/t<><74>%<25>*_<>V<EFBFBD><56><EFBFBD><EFBFBD><EFBFBD>Jj״֏^2<><32><EFBFBD>iM<69><4D>2<EFBFBD>Ys<59><73><EFBFBD>~<7E><>i<EFBFBD>/O<>L'<27><><EFBFBD>: /q<><71>%<25><>_<EFBFBD>V<EFBFBD><56><EFBFBD>Ƭ<EFBFBD><C6AC>kZ<6B>G/<2F><><EFBFBD><EFBFBD><EFBFBD><EFBFBD>yǡ<79>6<EFBFBD><36><EFBFBD><EFBFBD>>zɴ旧<C9B4><E697A7><EFBFBD><EFBFBD>Ѱ<EFBFBD><D1B0>% <20><>^2<>t|}^<5E><><EFBFBD><EFBFBD>F<EFBFBD>,<2C>y<EFBFBD> <1F>f^_<><5F>e`qJ<71>=<3D>Ǽ֏^3<><33> <0B><><EFBFBD>=<3D><13><>qEn <1F>f^_<><5F><EFBFBD><<3C>X<EFBFBD>Zs<5A><73><EFBFBD><EFBFBD><EFBFBD>k<EFBFBD><6B>uQ_K<5F><4B><EFBFBD>ﵕk^<5E>G<EFBFBD><47><EFBFBD><EFBFBD>e}<7D><>L<EFBFBD>A<EFBFBD><41><EFBFBD>Z?zͼ<7A>.<2E>kOw <1C><><EFBFBD>^<5E>y<EFBFBD> <1F>f^_<><5F><EFBFBD><EFBFBD><EFBFBD><EFBFBD>!<21><>s<EFBFBD><73><EFBFBD><EFBFBD>k<EFBFBD><6B>uq<75><71><EFBFBD><0F><>ژ֏^3<><33><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>@%<25><>]<5D><><EFBFBD><EFBFBD>k<EFBFBD>ԯ<EFBFBD><D4AF>^<5E>}<7D>7u<37><75><EFBFBD><EFBFBD><EFBFBD>k<EFBFBD><6B>uy<75><79>-<2D>8S <1F>ҽ<EFBFBD><D2BD>^3<><33><EFBFBD><EFBFBD><EFBFBD><EFBFBD>}<7D>Yz~<><7F>G<EFBFBD><47><EFBFBD><EFBFBD><EFBFBD>=4<>{<7B>s}<7D><><EFBFBD><EFBFBD>k<EFBFBD><6B>uy?<3F>r?<3F>ٱ<EFBFBD> &<26><><EFBFBD>k<EFBFBD>u<EFBFBD><75>2<EFBFBD>y<EFBFBD>O*=M<><4D><EFBFBD><EFBFBD><EFBFBD><EFBFBD>_vG<76><47><EFBFBD>9<EFBFBD><0E>I~<7E><><EFBFBD><EFBFBD><EFBFBD>̪qVu@<40><>? <1D>^<5E><><EFBFBD><EFBFBD>kf5<66>8<EFBFBD><38>N<>e2<65><32>2\<5C>=<3D>{<7B><><EFBFBD>><3E><>}֏^`<60>]<5D>{Q<>{<7B>:<1A>ε<EFBFBD><CEB5><EFBFBD>^0<><30>KqM<71><4D>{<7B>m<EFBFBD><6D>9<EFBFBD><39><EFBFBD> <1F>`b_<17><><18><>\h<> <1F>W<EFBFBD>~<7E><><EFBFBD>}]<5D>kb<6B><62>ۀ<EFBFBD>~<7E>vMl<4D><6C><13> AΉ<41><CE89><EFBFBD>M<EFBFBD>i<&<26>~<7E><><EFBFBD>}]s<><73><EFBFBD><EFBFBD><EFBFBD>2<13><>cb<63>G/<2F><><EFBFBD>UwM,w|~%<25><><EFBFBD>֏^0<><30><EFBFBD><EFBFBD><EFBFBD>X<EFBFBD>˓Z<CB93><5A><EFBFBD><EFBFBD><EFBFBD><EFBFBD> &<26>u<EFBFBD>]<13><>>!<21>N<EFBFBD><13><>֏^0<><30><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>.Ƕ<>^<5E>8<EFBFBD>9ɟk<C99F>w<08><>k<EFBFBD>&<26>~<7E><><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>Pg<50>.Ƕ<>^0<><30><EFBFBD><EFBFBD><EFBFBD>@6@h<><68>}^`Z?'<27><>١H<D9A1>1<EFBFBD><31>x<EFBFBD><78>G/<2F><><EFBFBD><EFBFBD><EFBFBD> Yz<59>Rl<52><6C><13>9<EFBFBD><39><EFBFBD>|<7C>ƶ<EFBFBD>^0<><30><EFBFBD><EFBFBD> :<3A><>6<EFBFBD>}<7D><><EFBFBD><EFBFBD><EFBFBD><EFBFBD>O<EFBFBD>x<EFBFBD>ƶ<EFBFBD>^0<><30><EFBFBD><EFBFBD><EFBFBD>w<EFBFBD><77><EFBFBD><EFBFBD><EFBFBD>L<><4C><EFBFBD>~<7E><><EFBFBD>6<EFBFBD>}<7D><><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>T<EFBFBD>T<EFBFBD>w)<29>~<7E><>i<EFBFBD><69><EFBFBD>O%Q<>Z<EFBFBD><5A><EFBFBD><EFBFBD> &<26>sR <1F><><EFBFBD><EFBFBD>m <1F>`b?'<27><>qJ<71>*8<><38><EFBFBD>֏^0<><30><EFBFBD><EFBFBD><EFBFBD>I<EFBFBD><49>*<2A><><EFBFBD><EFBFBD><EFBFBD> <0B><>sR<73><52>gEl<45><6C><05><>9<EFBFBD>ߪt<DFAA>:<3A>5<EFBFBD><35><EFBFBD>L<><4C>$~ <1D>J<EFBFBD><4A><EFBFBD> <0B><>s<EFBFBD><73>&)\<5C>(<28><05>~<7E><>i<EFBFBD><69><EFBFBD>g,
|
||||
V<EFBFBD><EFBFBD>P<EFBFBD><EFBFBD>^<10><>9I<0F><><EFBFBD>~<7E><><EFBFBD>`Z<><5A><EFBFBD><EFBFBD>?<3F><><EFBFBD><EFBFBD>`<12>'&q<><71><EFBFBD>p<EFBFBD><70><EFBFBD>/<2F><><EFBFBD><13><>a<14>W <1C>V~b J<>Cyҟ_0<5F><30><EFBFBD><EFBFBD><EFBFBD>C<EFBFBD>*<2A><><EFBFBD><EFBFBD>j<EFBFBD><6A>G<EFBFBD>=C<><43>~WO<57><1A><><EFBFBD><EFBFBD><EFBFBD><EFBFBD><7F>vϘ<76><CF98>e<EFBFBD>N<>G<EFBFBD><47><EFBFBD>0<EFBFBD> E.~Ô<><C394><EFBFBD>o<EFBFBD><6F>cwj<77><6A><EFBFBD><EFBFBD><EFBFBD>H<EFBFBD><48><EFBFBD><EFBFBD><EFBFBD> <1C>z<EFBFBD>$<24>><3E><>P<EFBFBD><50><16><> =<3D><>6ȷ<36><C8B7><EFBFBD> <1C><><EFBFBD><EFBFBD><EFBFBD>``<60><>&<26><><EFBFBD><EFBFBD><EFBFBD>0<EFBFBD> <1F><><EFBFBD><EFBFBD>)<29> m<>wL<77>)<29><>cJ?H<><48>
|
||||
S<EFBFBD>Q<EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><0F>tǔ~<7E><><EFBFBD>;<3B><>C<EFBFBD><43>ۧ<EFBFBD><DBA7><EFBFBD><EFBFBD><EFBFBD>1<EFBFBD> $ <0B><>)<29> O<>7L<37>G)¿cJ?<3F><0E> S<>Ab<41><62><EFBFBD><EFBFBD>r<>ǔ~<7E><0E>;<3B><>C<EFBFBD><43>ۧ<EFBFBD><DBA7><EFBFBD><EFBFBD><EFBFBD>1<EFBFBD> <1F>%<25><>)<29>(%<25>wL<77><07>ȿcJ?HD<48> S<>A<0E><><EFBFBD><EFBFBD>ҏÔ~<7E>y<EFBFBD>;<3B><><EFBFBD><0F>~s{<7B> <1D>><3E>|-<2D><>nx<6E><78>}6<><36>}B<>5<EFBFBD>Ϧ<EFBFBD><CFA6>O<EFBFBD><4F><1A><><EFBFBD><EFBFBD><EFBFBD> =<3D><>><3E><><EFBFBD>?<3F><>
|
||||
<EFBFBD>g#<23><>'<27>\{<7B>l<EFBFBD><EFBFBD><7F><EFBFBD><EFBFBD>n<EFBFBD>
|
||||
<EFBFBD><EFBFBD><EFBFBD><EFBFBD>S<EFBFBD><EFBFBD><EFBFBD>y<EFBFBD><EFBFBD><13><>_<EFBFBD><5F>'<27>\<5C><>l<EFBFBD><EFBFBD><7F><EFBFBD>j~<7E>
|
||||
<EFBFBD><EFBFBD><EFBFBD><EFBFBD>S<EFBFBD><EFBFBD><EFBFBD>Y<EFBFBD><EFBFBD>z<><7A>}6<><36>BO<42><4F><EFBFBD>&<26><>O詮<4F>٠<EFBFBD><D9A0> =<15>><3E><><EFBFBD>?<3F><>J<EFBFBD>gc<67><63>'<27><><EFBFBD>طO蹦<4F>ِ<EFBFBD><D990> =?<3F><><EFBFBD>><3E><>*<2A>g#<23><>'<27>T><3E>l<EFBFBD><EFBFBD><7F><EFBFBD>ꆟ
|
||||
<EFBFBD><EFBFBD><EFBFBD><EFBFBD>S<EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>z<>~6}B<>%<25>Ϧ<EFBFBD><CFA6>Oh<4F><68><EFBFBD><7F><EFBFBD> <1F>Y<EFBFBD>-<2D><>8<EFBFBD><38><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>K<EFBFBD><4B><EFBFBD>䥤<><E4A5A4>-<2D>4! 3Ĥ<33><C4A4><EFBFBD><z{<7B>_z<5F>t<EFBFBD><74>¹<EFBFBD><C2B9><EFBFBD>ۼ<EFBFBD><DBBC>o<EFBFBD><6F><EFBFBD>o1<6F><31><EFBFBD>n<EFBFBD><6E><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><7F><EFBFBD><EFBFBD>e<EFBFBD> M<>Ցf<D591><66>7<EFBFBD><37><EFBFBD>zp|۸w\<5C>/}p<>.<2E><><EFBFBD><EFBFBD><EFBFBD>pGϣ<47><CFA3> ;<3B><>WL!U<>}<7D>x<EFBFBD>'s8?<3F><>?<3F><>'m<>ֱk<><6B><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>u<EFBFBD><75> <1F><>/{<7B><><EFBFBD><14><><EFBFBD><EFBFBD><EFBFBD><7F><EFBFBD>&<01><>E<><45>6<EFBFBD>W~<7E><><EFBFBD><EFBFBD>/<2F><><EFBFBD>m<EFBFBD>/<2F><><EFBFBD><EFBFBD><EFBFBD>g<EFBFBD><67><EFBFBD><EFBFBD>@><07><>Y<EFBFBD><59>/<2F><>u<EFBFBD><75>Ι<EFBFBD>?<3F><07><><EFBFBD>x<EFBFBD><78><EFBFBD><EFBFBD>h |e<><65>0 <09><>R?<3F>'<27><><EFBFBD> <1C><><EFBFBD><EFBFBD><EFBFBD><EFBFBD>@<40><>=<3D>OG<4F>5<EFBFBD>N<c˼<63><CBBC><EFBFBD>a!f÷<66><C3B7><EFBFBD>-<<3C><>^<5E><>E<EFBFBD><45> <09><EFBFBD><C28F>:<3A><><EFBFBD>O~<7E><><EFBFBD> m<><6D>J<EFBFBD>#<><7F><EFBFBD>W<EFBFBD>h<EFBFBD>x<EFBFBD><78>(<28>-<2D><>U?zb<7A>H<EFBFBD><48><EFBFBD><EFBFBD><EFBFBD>W<EFBFBD>f><3E><><EFBFBD><EFBFBD>J߹<4A><DFB9>>g<><67><EFBFBD><07><><EFBFBD><EFBFBD>7<1B>+\<5C><>'?<3F>/<2F><>'^<5E><<3C><><EFBFBD>i<EFBFBD><69>~<7E>o¬ <18>ٶ<EFBFBD><D9B6><EFBFBD>OV<4F><56>^<5E><><EFBFBD><EFBFBD><EFBFBD><0F><>7{<7B><>1<EFBFBD>g?<3F> <>O6)<29>~rPұ<50><D2B1><EFBFBD><0F><><EFBFBD>Q<EFBFBD>O<05>GYIv.<2E>q<EFBFBD><71><EFBFBD><EFBFBD>i<EFBFBD><69><EFBFBD>u
|
||||
/<2F><><EFBFBD><EFBFBD>$<24><>k<EFBFBD><6B> 꺾0<EABABE>G)<29>:<3A><><EFBFBD>t}9U`{<0F><> <1C><><EFBFBD>#<23><><EFBFBD><EFBFBD>F<EFBFBD>Ol<4F>m
|
||||
߹<15>+ <0B>)|<7C>V\<5C>
|
||||
<EFBFBD>)|<7C>V|w<>m<EFBFBD><6D>wH<77><03><><EFBFBD><EFBFBD><EFBFBD>o<EFBFBD><6F>O}<7D>֙<EFBFBD><D699>ݗ<><DD97><EFBFBD>qn<71><6E> [<5B>}eJO<4A><4F><EFBFBD>)<29><><EFBFBD><EFBFBD>W<EFBFBD><57><EFBFBD>en<65><6E> <0B><>}eJO<4A><4F><EFBFBD>)<29><><EFBFBD><EFBFBD>W<EFBFBD><57><EFBFBD>Yn<59><6E> [<5B>}iJ{O<>}J<>k.<2E>%<25>h<EFBFBD>&<26>M<EFBFBD><4D>m<EFBFBD>~J[<5B><><EFBFBD><EFBFBD><EFBFBD><EFBFBD>S3<53>u.<2E><>j<EFBFBD><6A>)=<3D><>ۦ<EFBFBD><DBA6>~v_<76><5F>S<03>mJ/<2F>d<EFBFBD><64>)=<3D><>ۦ<EFBFBD><DBA6> v_<76><5F>SӺmJ/<2F>^<5E><>)m<><6D><EFBFBD><19><>o<EFBFBD>W&<26>Ԩn<D4A8><6E> ;<3B>}iJ{<7B><>}J<><4A>U<EFBFBD>6<>:<3A><><EFBFBD>!<21><><EFBFBD><EFBFBD>3ye<79><65><EFBFBD>L<EFBFBD><4C>/<2F>6<EFBFBD>6<><36>ʔ<EFBFBD>:<3A>mSzak<61><6B>L<EFBFBD><4C><17>6<EFBFBD>6<><36>ʔ<EFBFBD><CA94><EFBFBD>mSza;<3B><>Li<4C><EFBFBD><7F><EFBFBD>u<EFBFBD><75><EFBFBD>4<EFBFBD><34><EFBFBD><EFBFBD>><3E><EFBFBD><D7B5><EFBFBD><1B><>]<5D><><EFBFBD><EFBFBD><EFBFBD>~ʳ<><CAB3><EFBFBD>{=[<5B>-<2D><><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><17><><EFBFBD>lo<6C><6F> <0B><>}aBOm<4F><6D> <09><><EFBFBD><EFBFBD>&<26><>`o<><6F> ;<3B>}aBO<42><4F><EFBFBD> <09><><EFBFBD><EFBFBD>W&<26>7<EFBFBD><37>&<26><><EFBFBD>z_<7A><5F>m<EFBFBD><6D>6<EFBFBD><36><EFBFBD><EFBFBD><EFBFBD>}ɍ<>w<EFBFBD><77>v<><76><EFBFBD><EFBFBD>}<7D><><EFBFBD><EFBFBD>Sz/D<>f<0F><><EFBFBD><EFBFBD>f<EFBFBD>ܚ<>i<>S<EFBFBD>~<7E><>'<1A><><EFBFBD>߭<EFBFBD> y<>><3E><>wk~B<><42>O<EFBFBD><4F>ݚ<EFBFBD>f<>S<EFBFBD>~<7E><>'TOS<4F>v<EFBFBD><76>'$<19><><EFBFBD>߭<EFBFBD>E<EFBFBD><45><EFBFBD>|s<>owkns<6E>
|
||||
n<EFBFBD>O<EFBFBD>9<EFBFBD>)}<7D>[<5B>^<5E>}J<><4A><EFBFBD><EFBFBD><EFBFBD><EFBFBD>c<EFBFBD><63><EFBFBD><EFBFBD>5?<3F><>ا<EFBFBD><D8A7>n͏, O3<4F>n<EFBFBD><6E>'<27><1B><><EFBFBD>߭<EFBFBD> <09><>Ӕ<EFBFBD>ݭ<EFBFBD><DDAD>#l3<6C><33>|<7C>ݭ<EFBFBD><DDAD><EFBFBD>7<EFBFBD>5?<3F><>ا<EFBFBD><D8A7>n<EFBFBD>O<EFBFBD>><3E>)}<7D>[<5B><13><>}J<><4A><EFBFBD><EFBFBD><EFBFBD><EFBFBD>c<EFBFBD><63><EFBFBD><EFBFBD>5?<3F>z<EFBFBD>3<EFBFBD>v<EFBFBD><76>'<27> O3<4F>v<EFBFBD><76>'T <1E><><EFBFBD>߭<EFBFBD>eI<65>s<EFBFBD><73>,<2C>'<27><>><3E><>wk~<7E><10>f<EFBFBD><66>n͏<6E>!ۄ<>߭<EFBFBD><DFAD>7d<37><64><EFBFBD><EFBFBD>5?2<>l<13>~<7E><>G<EFBFBD><47>}B<><42><EFBFBD><EFBFBD><EFBFBD><12><>Ͼۭ<CFBE><DBAD> <1F>ɝ<EFBFBD>;t<><74>ú<ܚ_/<2F>SK<53>R<EFBFBD>L閰<4C>T+֎R?<12>{ʙ<><CA99>A傹^<>H<EFBFBD> <1F><>s<EFBFBD>{<7B>9}<7D><><EFBFBD>\<5C><><17><><EFBFBD><EFBFBD>L<EFBFBD><4C><EFBFBD><EFBFBD><EFBFBD><EFBFBD>uy<><17>I~a<12><><EFBFBD><EFBFBD><EFBFBD>|2<><EFBFBD>Vz<56><7A><𗣔<><F097A394><EFBFBD>{<7B><><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>?<3F><>7<EFBFBD><37>?<3F>ß<EFBFBD><C39F><EFBFBD>Pz<50>CK<43>^<5E><18>|ϫ<><CFAB><EFBFBD><EFBFBD><EFBFBD><17>'<13><><EFBFBD><EFBFBD>'3<><33><EFBFBD><EFBFBD>OVF<56>38O\#<23><<3C>ẇ<77><CC87><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>+_<><5F><EFBFBD>b.<2E>h<EFBFBD>u<EFBFBD><75><EFBFBD><EFBFBD><EFBFBD><EFBFBD>e<EFBFBD><65>)<29><><EFBFBD>Z<14>s U<><55><EFBFBD>K<EFBFBD><4B>{<7B><><EFBFBD>~<7E>{<7B><><EFBFBD>f<EFBFBD><66><EFBFBD><EFBFBD>T<EFBFBD><54><EFBFBD><03><05><>N<<3C>2<EFBFBD><32><EFBFBD><EFBFBD><EFBFBD>dR<64><52><EFBFBD>q<EFBFBD><71><<3C>^?<3F><>6<EFBFBD>9 -uͰ%h\<5C>Q+<2B>ɻ<EFBFBD>E<EFBFBD>(<28>w#<23>4><3E>F<EFBFBD><46>i<EFBFBD><69><EFBFBD>1<EFBFBD><31><EFBFBD>q~<7E>4<EFBFBD>=c<>M<><4D>4><3E><><EFBFBD>5Ӡ<35><D3A0><EFBFBD>c~<7E><16><>&#`<60>(<28><><EFBFBD>,>i<><69>kf<6B>գ-<2D><><EFBFBD><EFBFBD>g<EFBFBD><67><EFBFBD><EFBFBD><EFBFBD>L<EFBFBD><4C>R<EFBFBD><52><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>ï<EFBFBD>Em<45><04>4<EFBFBD><34>֢<EFBFBD><D6A2><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>:<3A><><11>c<EFBFBD><1B><17>ea=<3D>8O<1A>J<EFBFBD>i<EFBFBD><69>,<2C><><EFBFBD>#<23>{[ۗF<DB97>n<7F>BO<42>I_<49><05><>Yn3ٝq[o<>W:<3A><><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>k<EFBFBD><6B> s<>óݦ<C3B3>[<5B><02>Y<EFBFBD>=c<><63><EFBFBD><7F><EFBFBD> <1F><>/Z<><5A><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>a<16><>G<0F>d<EFBFBD><64><EFBFBD><EFBFBD><EFBFBD><EFBFBD>P<EFBFBD>G <1E><>)<29><>~<7E>~ r<>觠<EFBFBD>5j<35>d/"c 6s<36><73><EFBFBD>Y:,<2C>جgKǴ1<C7B4><31><EFBFBD><EFBFBD><EFBFBD><EFBFBD>*<2A>t<04><><02><>r`Pj<50><6A>-<2D>1<EFBFBD><01><1A><>f<EFBFBD>jbت<62><D8AA>3<EFBFBD><18> <1F>X<EFBFBD>M<EFBFBD><4D>x<1A>=3Ƶ<J<01>p<EFBFBD>T)~<7E>1<EFBFBD><31>B<EFBFBD><42><01>O<EFBFBD><4F>a<> \<5C>j.N<>/<2F>
|
||||
k|<7C><07><18>傱>R <1F><><EFBFBD><EFBFBD>c<EFBFBD><63><EFBFBD>P<01>Q<EFBFBD>Ÿq<C5B8>K<EFBFBD>We<57><65><EFBFBD>4<EFBFBD><34>4vB<14> <1E><> <1D><><EFBFBD>8h3<68><05>Mh<4D><68>q<E281AE>d<EFBFBD>?b<>qϝ́j<<3C>̅:H`|<7C><><EFBFBD><EFBFBD>Xme(<28>.ᯍ<> G<><47><EFBFBD><03>O~C
|
||||
<EFBFBD><EFBFBD><EFBFBD>(<28><><EFBFBD>a<1B>cz<63>'[<5B> Gg <0B>V<EFBFBD><56>30<33>hv<><76>[>Oϳ<4F>K륍 <1C><><EFBFBD>:@X<>k<EFBFBD><6B>1r<31>Kx{<7B>4<EFBFBD><34>I<EFBFBD><49><EFBFBD>q<EFBFBD>}<7D><><EFBFBD><EFBFBD><EFBFBD>i'<27>m<EFBFBD><6D><EFBFBD><EFBFBD><EFBFBD>LK<13><><EFBFBD>e<EFBFBD><65><EFBFBD>k<><6B><EFBFBD><06>
|
||||
,Ɲ<>ɓIh<49>[<5B>
|
||||
<EFBFBD>EL<EFBFBD>,X<>ᯜw<<3C>Y<EFBFBD>7S<18>F<EFBFBD><46>
|
||||
M<><07>W<><57>1+<10><>b<EFBFBD>ƸR<C6B8>`W<>RҽT <1C>\<5C>q<EFBFBD><71><EFBFBD>%<25>R[O <1C><12><><06><0E><11>[<5B>m<1B>P<EFBFBD> <1D><><EFBFBD>N<EFBFBD>x-<2D><><EFBFBD><EFBFBD>xN-u)<29><> ,Ƶc<>K<EFBFBD>><3E><>\<5C>݉0< @mk0<> <0C>K<EFBFBD>q<EFBFBD><71><EFBFBD><EFBFBD><EFBFBD>l9Ke<4B><65>k<EFBFBD><6B>מ<EFBFBD>g<EFBFBD><67>k<EFBFBD><6B><14><>3`<60><06>K<EFBFBD><4B><EFBFBD>;y<>g<EFBFBD><67><EFBFBD>Lē<4C> <1E>f<EFBFBD>rz.r<08><06>7<EFBFBD><37><EFBFBD>1<03>)<29><>l <1E><15>4<EFBFBD><34><EFBFBD><EFBFBD>=0<07><><EFBFBD>M/<2F>o+<2B>j<EFBFBD><6A>tK<74>[<02>#<23><><18><>!<21>R<EFBFBD>i<EFBFBD>'<27>5^щ<><D189>1yaP<61>`);<03><><EFBFBD><EFBFBD><EFBFBD><EFBFBD>8<EFBFBD><38><EFBFBD>H<EFBFBD>9<EFBFBD>ã<EFBFBD>\
|
||||
<EFBFBD>4<EFBFBD><EFBFBD>v<EFBFBD>8Lz<EFBFBD><EFBFBD>g<EFBFBD>=<02><><EFBFBD><EFBFBD>8<EFBFBD>mVK *V<> <09>ԘK<19>I<EFBFBD>\`#<23>v<EFBFBD><76>%<25>B<EFBFBD><42>#Ж<><<3C>
|
||||
<EFBFBD>x,<2C><<3C><06>T<EFBFBD><54>c<EFBFBD><63><03>of <09>c* .Ga<47>h8<68><38> <1E>z<EFBFBD>r" ѱ<><D1B1> <1C>01Ib<>~@`<02>q<EFBFBD><16><02>ĠDtț7!L <20><><EFBFBD><EFBFBD>p<EFBFBD>uW=<3D>@<40><>x<EFBFBD><78>m<13><19>fLM<4C><4D><EFBFBD> ;<3B>Q<EFBFBD>z<EFBFBD>Y<EFBFBD>E<EFBFBD><45><EFBFBD>ʽ&<26><><1A>[aㇳxX<78>ы_<D18B>ۍJ/ޜ<>x<EFBFBD> 5n7z *<2A> <1F>ڽ <1D><>-<2D>Xa<58>o<1A>ؐ<EFBFBD><16><>ts瀖<7F><E78096><EFBFBD><EFBFBD><18><><EFBFBD><EFBFBD><EFBFBD><EFBFBD>:<3A><17>w 6bO<>ˁ<EFBFBD> Ru<52>4t<34><74><EFBFBD><04><18>@<40><>sa <1D><>QZ<51><1B>s3<73>;;<3B>
|
||||
M<EFBFBD><EFBFBD>e.l <1E><>ޝ <0B> n<><6E><EFBFBD><EFBFBD>n<EFBFBD>m ;u<>B<EFBFBD>3`e<>i<05>C^Q<><51>]aGF<><46><15> <1C><>ư<EFBFBD>
|
||||
t<EFBFBD> <1C><>ԽU<02><><EFBFBD><EFBFBD><EFBFBD>!<21>&v<18><> <20><><EFBFBD>ӱ?T<>yİ<79><C4B0><EFBFBD><EFBFBD>w:<3A><><18>o<EFBFBD>I<15>3<EFBFBD>x
|
||||
<EFBFBD>];<3B><>8 P,<2C><>P lч<6C>3b\<5C> <09><> <1F><>mPTƈq<C688><71>)M<02><>|<7C><><EFBFBD>q<EFBFBD><71><EFBFBD>3 <0C>+(<28> <1D>'<27><><EFBFBD>r)<29><><EFBFBD><EFBFBD><EFBFBD>,<2C>N<EFBFBD><4E>R<EFBFBD><52>6<><36><EFBFBD>yz\K<>3!<21><><EFBFBD>[<5B><>,(s<>Z$<24><><EFBFBD><14>$<24>0\<5C>=<14>Q<EFBFBD><51> X<><58><EFBFBD>(<28><>XǛ<58>8<EFBFBD><38>p
|
||||
B<EFBFBD><EFBFBD>T <20>`<60><01>qV9<56><39><EFBFBD><EFBFBD>b<>
|
||||
<EFBFBD>W<EFBFBD>D<EFBFBD>p
|
||||
E<EFBFBD>H4<EFBFBD>W<EFBFBD><EFBFBD><EFBFBD><EFBFBD>2 <20><>
|
||||
!X<><58><EFBFBD><EFBFBD>!<21><><EFBFBD><EFBFBD><EFBFBD> <05><>>& a<>{<7B><>ص#54<35>A<EFBFBD>
|
||||
h}<7D><>0#l o<> a<><61>\y[ <0B> i<>ߛVq<><71><EFBFBD>AD'_<>@p<>q)<29><>p<EFBFBD><70><EFBFBD>c<>O|{ ;<3B>|<7C><08>|<13><>o<<3C>MH <1C>D`cyn<79><6E>S*!H*<2A><><EFBFBD><EFBFBD>;0?<3F>(<0F><>C9<43><39><1B>1<EFBFBD><31>G@<40><><EFBFBD>$)<29>AT`c<>+!<21><>r3<72><33>QOȻX<11>)ȶ<12>q`<60>p{x<><78>p<EFBFBD>03<30><33><EFBFBD>?<3F><18>th_2#nTA<54><41><0E>}<7D>0<EFBFBD><30>g<1A>6<EFBFBD><36>l<08>.<05><><EFBFBD>h<EFBFBD><68><EFBFBD><EFBFBD><EFBFBD><EFBFBD>;<3B><03>
|
||||
<EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD> <09>)<29><>#<23><>0
|
||||
R轳q<EFBFBD>l<><14><><EFBFBD><EFBFBD><EFBFBD><EFBFBD>̦<EFBFBD><CCA6>i\1 <0C>Y[<5B><><15><>T<14> <0C>#Q<> <1F>`4<><34><EFBFBD>)<29>mN7N<37>n^bE<>/<2F>^<5E>ݿ'?3<>a<1A><><EFBFBD>Ep<45><70><EFBFBD><EFBFBD><EFBFBD>\<5C><><EFBFBD>,,T<>30-<2D><>+iQ<69>flPBB<42><42><EFBFBD>J<EFBFBD>9<19>6qϏ<71><CF8F><EFBFBD><EFBFBD>ǁ<EFBFBD><C781>g <0C> vM<76>a8<02><>g:p <00><>1<0E> <17><>%j<<3C>Fx<46><78>qsJ\fR<66>SW<53>5<EFBFBD>aP9k
|
||||
³<>~5c 'S)<29>x<EFBFBD><78><EFBFBD><EFBFBD>f<EFBFBD><66><EFBFBD>v<EFBFBD><76>.+<2B><>qI<71>ӕ<04><><EFBFBD>U7 1<>k<EFBFBD>(<28>a<EFBFBD><0F>aU<61>es<65>1nrq<72><71> <1E><>=SK<53>0,..<2E><>y<EFBFBD>z<EFBFBD>TrK o<><6F>Ѭ<05><>qd<71> ,q<>
|
||||
3^<11><><EFBFBD>m<EFBFBD>d<EFBFBD><64>k5<04>(<28><>30hX<68>{<7B>\|<7C>x]h<> aE <0B><>C<EFBFBD>Yl;6p<36>-<2D>5<EFBFBD><35>CF1ڪ<31><DAAA>1n<31>*<2A><>b2Mq<4D>h}
|
||||
1<EFBFBD><EFBFBD>4 <0B>LS<4C><53>%f /2<07>K 4<><34><EFBFBD><EFBFBD><EFBFBD><EFBFBD><04>/1H<31>_9h<04><><EFBFBD><16><13>9<<3C>A ;<3B><0E><>q<EFBFBD><71><EFBFBD>2<EFBFBD><08><>n<EFBFBD><6E>3<EFBFBD>A<EFBFBD>C <1F>T:<3A>DX>3ƍ|F*<2A><><EFBFBD>`EC<0F><EFBFBD>wj<77><6A>L_܀<5F><05>0v<30><76><EFBFBD><18><EFBFBD>;$ɰ<>i<EFBFBD><69>1<10>9XT<58><54> lP'=<3D><>f<EFBFBD>Gn4cT
|
||||
p=<3D>R{<7B><> A Î<>oͳƸN7<4E><37>喇 <09><><EFBFBD> <1E>y<EFBFBD>d<EFBFBD>s<EFBFBD><73><EFBFBD>c #<23> <1C>d <0C>
|
||||
<EFBFBD> ]ƠO !\N:<3A><18>K<EFBFBD><4B>S<EFBFBD>A <0C>o,~<7E>IY<49>U<EFBFBD>ڋ<EFBFBD><DA8B>N<EFBFBD><4E><EFBFBD>iՋ<69>1<EFBFBD><31><EFBFBD>;<3B><>7 7
|
||||
<EFBFBD>AS<EFBFBD>O<EFBFBD>KX|h :E<><45>y<17>+b<>6[L<> <20>:^p 1n`<60>t<EFBFBD>>`<7<>w<EFBFBD>Ö<EFBFBD><14><>`<60>7D<><44><EFBFBD><EFBFBD><EFBFBD><EFBFBD>i<EFBFBD><69><EFBFBD><EFBFBD>`<60><><EFBFBD><EFBFBD>Ģ<EFBFBD><C4A2><EFBFBD><EFBFBD>2<EFBFBD><32><EFBFBD>2>=<3D>j;mE <1B><><EFBFBD><EFBFBD> <1E>f<02><><EFBFBD><EFBFBD><EFBFBD>a<07><>Y<EFBFBD><59>܆<02>%{<18>d<EFBFBD>90n<30>f<EFBFBD><66><EFBFBD>^<5E>6<EFBFBD>q<EFBFBD><71><EFBFBD>Rl<52><6C><EFBFBD><EFBFBD>]<5D>7<EFBFBD>eM<7F><4D> U<>`<60>N9<4E>p,<2C><>!<21>w<18>8687<38><37><EFBFBD><EFBFBD>ɶ<EFBFBD><C9B6>A\U<><55><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>O<18>U<EFBFBD> M<>zO<7A>gЇ<67>h<EFBFBD>a<EFBFBD>v<EFBFBD>vƠ<76><C6A0><EFBFBD>7,q^<5E>-u<>æ<EFBFBD>W@ <1D><07>xA$ d<>G`:6W<36>ش-' <0C>D><3E><><EFBFBD><EFBFBD>CAY<18>%<25>O<>1\<5C><><EFBFBD><EFBFBD> <09>N <10><><EFBFBD>g<EFBFBD><67><13>8l<38>Ƒƺ<C691><C6BA><EFBFBD>U<EFBFBD>~<7E><><EFBFBD><EFBFBD><EFBFBD> <00><13>%6<><36>!b<><62><EFBFBD><EFBFBD>=<3D><>ݩ<1B>H<13>[jz<6A>W*xX(d<>5\5<> <1F><>Kg<4B><67>e<EFBFBD>^<5E><> <1E><>+<2B><>j oƢGMH悻<48>?<3F><>M<EFBFBD>(V8<56>+v<><76>0<EFBFBD><30>g<EFBFBD><67> m^<15> z<>W<EFBFBD>h<18><>Wm<57><6D><03><>;<16>y,89<38> <1E> <0B>R<EFBFBD>(.p?<3F>X<EFBFBD><16>/0<><05>CcQ<63><51>z<EFBFBD>2<EFBFBD><32>6<EFBFBD>V<EFBFBD>bQ:<3A><>P<EFBFBD>`iS<69>Kez<18>F;<3B>n<12>K<EFBFBD>&<= & <1C><>E <1C>p<EFBFBD>PLdx<64>$<24>]<5D>+<<8 Y<08>?<3F>?"6)<29>mMW<4D>bb<62>Z<EFBFBD>4<EFBFBD><34><EFBFBD>f/R}+<2B><> <12><><EFBFBD>J<EFBFBD>I<EFBFBD> M\<5C>Ղ <1F><><EFBFBD><EFBFBD><EFBFBD>axc<10><><EFBFBD><EFBFBD><EFBFBD><EFBFBD>c<EFBFBD>A<EFBFBD><41><EFBFBD>ԯ<EFBFBD><06>
|
||||
[= <0C>|b<><62><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>b<EFBFBD><62> <1E>`<T;J`8<><38><EFBFBD>A" <03><><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>FlEl}:ㄵ3_<33>Fb<46>~<7E>"Fwᛰʌ-<2D>I<EFBFBD>8쓇<38>I <1C>>Y<><59><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>lp\<10><>W<EFBFBD>r臽f<>s<EFBFBD>;<3B><><EFBFBD><EFBFBD>c<>"<22><><EFBFBD><EFBFBD>/<2F><><EFBFBD>m<EFBFBD><19><><EFBFBD>%ˊ<>}<7D><><EFBFBD><EFBFBD>u<>M<EFBFBD><10><><EFBFBD>6!6<18><><EFBFBD>9<EFBFBD><39><EFBFBD>뜀F<EB9C80>.s <1C>3<EFBFBD>2hu<68><75>g<EFBFBD><67><EFBFBD>Tw<05>f<><66><EFBFBD>b
|
||||
'k<><6B><EFBFBD><EFBFBD>)$<24><><EFBFBD>W~ZJ*<2A><><EFBFBD><EFBFBD>34a<34><61>7< <0C><01>دn<D8AF>4<EFBFBD>鞄<EFBFBD><E99E84>
|
||||
Ma<> <1C>2OE*<2A><><EFBFBD> <1E>*<2A><>:T9<<0E>.<2E><><EFBFBD>"И<><D098>A%h<><68>sa<73><03>B<EFBFBD> <1E><>?<3F>Ou<4F><13>Syxܤ<78>M<EFBFBD><4D>ƐE<08><>l z<>
|
||||
F8`ke<6B>&<26>whfZM<5A><4D>C|<0E><><EFBFBD>K<EFBFBD>N<EFBFBD><4E><EFBFBD>u M<><1B><1B><><EFBFBD><18>
|
||||
|
||||
8da<EFBFBD>T<><54><EFBFBD>I<EFBFBD><49>t<15><>6<EFBFBD> \<5C>a2iDy<44>T<EFBFBD>2/ <0C> ֥x VsB<73><42><EFBFBD>Q<EFBFBD>N.<2E><>as4i<34><69><18>.<2E>Fc<46>+Cl<43>)<29><><EFBFBD>DzR<10>$}<7D>u<EFBFBD> <1E><><EFBFBD><EFBFBD>! S<> 6U<36><55>><3E><><EFBFBD><EFBFBD><11>Z<12><> H*<2A>U<EFBFBD>$A<>]5H \ <1F><>S<EFBFBD>aP<61>p<EFBFBD><70>͞ 8<><38><EFBFBD>8?i <0C>V<EFBFBD><56>U8<55>NS
|
||||
<EFBFBD><16><><EFBFBD><EFBFBD>ދ<EFBFBD>qs<71>b<EFBFBD><62>
|
||||
~<7E><><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><06> krư<72>RGB<47><42>I<EFBFBD><0E>̏q<CC8F><71><EFBFBD><EFBFBD>)<29><05><EFBFBD><7F><EFBFBD>U+<2B><>v%<25>8<EFBFBD><13> <1F> <1E>P<>S<EFBFBD>8,l<><6C>O0<4F><30>tx\Ý<>`)1nU<6E><55>$<24>X<>K<~<7E>ፂ<EFBFBD><14><><15>M <18>'m<>xw4Mg<4D><67>0<EFBFBD><30><EFBFBD>zH<7A><48>^)P<>G
|
||||
<EFBFBD>ރj<EFBFBD><͛<><CD9B><05><08>=<04> <1D><><EFBFBD>*$<24><>D<03>{<7B>h;<3B><>Tb<54><05><><EFBFBD><EFBFBD><EFBFBD>y<>!<05>$wNf<12>Պ<EFBFBD><D58A>%<25>Z<EFBFBD>Ĥ><3E><xH<16>UN<55><04>!<21><0F><>$[<5B><>HZ<11><01><><EFBFBD><EFBFBD>%<25>`B<04><>
|
||||
<EFBFBD><EFBFBD> <1D>-I5<49>ct<03><>8=.<2E>-<2D> <18>.<2E>aɉ7 <1E><><EFBFBD><EFBFBD><EFBFBD>N- <09><12><><EFBFBD>ڀ 0<>"<22>
|
||||
<EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>g<><67><EFBFBD>
|
||||
<EFBFBD>[<5B>+~j<>6<EFBFBD><06><>q<EFBFBD>\<07>pESX<53><58>T+N<><4E>.w>_zGSC7 1<>8<EFBFBD>A<EFBFBD>8<EFBFBD><38><EFBFBD>>E<><45> <0F><15>u<EFBFBD> όl4<18>a<EFBFBD>t<EFBFBD>yJO#<23><> K*/Nj<19><><EFBFBD>4<EFBFBD><13><><EFBFBD>V<EFBFBD><56><EFBFBD>h<>4Ó61<36>1<EFBFBD> 7<><37><02><>ê$<24><><EFBFBD>z<EFBFBD><15>3<EFBFBD>=bQ<62>ϋ<EFBFBD>
|
||||
2 yΛ<>E<EFBFBD><01><><EFBFBD> <1D><><1A>劇:<05><><EFBFBD>E#<23><><EFBFBD>&p<><70><08>9<EFBFBD> <1F>q
|
||||
<EFBFBD><EFBFBD>C<EFBFBD><EFBFBD>@F<><07><>W<EFBFBD> <20><><EFBFBD><EFBFBD><EFBFBD><0E><><EFBFBD>E<1B><> JX<58><C28F>S<EFBFBD><53>#,Tz3-χFbt<>A<EFBFBD><41>sƛ<73>uilL<6C>[i<>1<EFBFBD><31>q<EFBFBD>J<EFBFBD><4A><EFBFBD>ʾq<CABE>p>.ߟ<>{Ll0?<3F><><EFBFBD>%d z nJ<6E>
|
||||
V<EFBFBD>76<EFBFBD><EFBFBD>¨ce<EFBFBD> Oa<4F><05>y<10><>5<EFBFBD><35><19>K<EFBFBD><4B><EFBFBD><EFBFBD><EFBFBD><08>4<EFBFBD>S#<23>A<EFBFBD><41>jMYH<59>D<EFBFBD><44><EFBFBD>M<EFBFBD><4D><03><>Gg`#<23>mƦ<6D><C6A6><EFBFBD>|<7C>R<EFBFBD> VvC<76><43><EFBFBD><EFBFBD>V$<24><>ݡȳ0<C8B3>IO<49>Z'<27><16>[,ܝ<><DC9D><EFBFBD>m<EFBFBD><6D><06>2-<2D>M9<4D>(iq@
|
||||
<EFBFBD><EFBFBD><EFBFBD>, kL2D
|
||||
a<EFBFBD>c<EFBFBD><04><><15>5<EFBFBD>{<7B><0E><><1B>k<EFBFBD> *<2A><><EFBFBD><EFBFBD><1B>!x4G]
|
||||
A<EFBFBD>aM<EFBFBD><EFBFBD>%*<2A><>an<61>B <1E><02>>(9<><14>̀<EFBFBD><CD80><EFBFBD>1<EFBFBD><31>)t<><74>uy<75><79><EFBFBD>1d 2_<32>L:<3A><>d<EFBFBD>7ܙ= <00><><EFBFBD>8<EFBFBD>.<%<25><1A>&<26>aQV3<56><33><EFBFBD> <1E>a
|
||||
<EFBFBD>T<EFBFBD>l(<28>y<04><><EFBFBD><EFBFBD>1(<28><>Y,<2C>*<2A><1A>^<1B>;<3B>~G<>B6xx[<5B>i2o)<29><><EFBFBD><EFBFBD>:<3A><>`<60>1<EFBFBD><31><EFBFBD><EFBFBD>
|
||||
:<3A>e<EFBFBD>ր<EFBFBD>A<EFBFBD><41> &<26>l<><6C><EFBFBD> '<27><>o<EFBFBD>hq<68>T2<54><06>{<18><><EFBFBD>3t<33>P<EFBFBD><50><EFBFBD>3<19><><EFBFBD>N,1ؒ|
|
||||
*M<>p; t<>`q<><71>Ac<41>J<EFBFBD><4A><1A>7<EFBFBD>3<EFBFBD><33>0<EFBFBD>!<03>㨙0<E3A899> f¸l˪5<CBAA> <1E><> <1C> l<>;<3B><><EFBFBD> =<3D>$<24><><EFBFBD><EFBFBD>Gf<12><>^<5E>C<EFBFBD> <19><><18>L<08>0:<3A>f<EFBFBD>a<EFBFBD>q<EFBFBD><71>
|
||||
<EFBFBD><EFBFBD>t<EFBFBD>S˛<EFBFBD><EFBFBD>Y`<60><>9<EFBFBD><39>Ď^<5E>+ <20><>
|
||||
|#<23><><EFBFBD>T<EFBFBD>+%<25><>ِ<>=<3D>? .<2E>ˋC <0C><>d<EFBFBD><64>F%ﲗ*G/v(t<><74>o<EFBFBD>j1<6A>k<EFBFBD><6B>od`V:<3A>xT<78>!<21>DV-<2D> <1C>娅7<E5A885><37>G><03> <09><>4<EFBFBD><34><EFBFBD><EFBFBD>^<5E><>gc\҅<>E0 r<><72> °<06>Z<EFBFBD><5A>`><3E> <18>UD<55>+<2B>TA* <1D><><03><><EFBFBD>d<EFBFBD>$<24>i<EFBFBD>8<><38><EFBFBD>=R<>8<0E><05><gY[<5B>p<EFBFBD><70><EFBFBD>W<EFBFBD><57>x<EFBFBD><78>"<22><>u<EFBFBD>5<EFBFBD><35>%o<15>2
|
||||
<EFBFBD><EFBFBD><EFBFBD><EFBFBD><19><><EFBFBD><EFBFBD>v<>9<EFBFBD>-C<><43><EFBFBD><EFBFBD>3<EFBFBD>Q<EFBFBD>(<28><><18><>=Oo<4F>θ<06>?<3F> hXRˮγP_/̒<><CC92>F<EFBFBD><46><EFBFBD><EFBFBD><EFBFBD>E<EFBFBD><45>[<5B>ZZ<5A><5A><13><><EFBFBD>\<5C> <1E><> <1E><>2<EFBFBD><32>N<EFBFBD><4E> 1ww~<7E><><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>&S61VN+]<5D> ˡΠF <1D><><19>:H<><48><EFBFBD>#<23><><EFBFBD><EFBFBD>FLq<4C><71><EFBFBD>E<EFBFBD>;<3B><><EFBFBD>E<EFBFBD>t<EFBFBD>B<EFBFBD><42><EFBFBD>_cP#2<>胀"JB<4A><42>z<EFBFBD>E<EFBFBD>o<EFBFBD>Wb<01><><EFBFBD>j'moh<6F><68>8<EFBFBD><EFBFBD><7F><EFBFBD>SG<53><47>P<EFBFBD> <00><><EFBFBD>0N <00><>h| z#<18><>њ<EFBFBD>'a c01<30>\z<><7A>Ժ<>*9_3<5F>1Jl5<6C><35>Y<EFBFBD><18>6<EFBFBD><36>-<2D><>M<EFBFBD>%<25>v9<><7F>Ɵ3<C69F><0E>VMG<4D><47><EFBFBD><EFBFBD>><0E>.z<> <20>q+<2B><><0F>4<EFBFBD><34>de<64><65><EFBFBD><ۥ<><DBA5><EFBFBD>qv <09><1B>jK$nw\<5C><>b <1C>Bř<42>Jl `0<>a<EFBFBD><61>F/
|
||||
<EFBFBD><EFBFBD><EFBFBD><13>P<EFBFBD>'<27>u:<06><><EFBFBD><EFBFBD><EFBFBD>,<2C>i0-<2D><><02>3H<33>L<EFBFBD>p<EFBFBD>Y<EFBFBD> <0C>!<21><><EFBFBD>|T<><54><18>Rq<52>ڤ.5<EFBFBD><EFBFBD>=).p<>&<26>-<2D><>B.N<><4E>RO<52>4p0$ <20><>g<EFBFBD>[~<7E>1
|
||||
<EFBFBD>Wt<EFBFBD><EFBFBD>;ۘ<><18>Ĝ<EFBFBD>d`<11>DƂ<44><C682><EFBFBD>6CZ<43> Ogp<67><70>rZ<07>ԓ<EFBFBD><D493>DK<>z><3E><>̩<EFBFBD>a]p<><70>1}<7D><><EFBFBD><EFBFBD>b<EFBFBD><07>ש&E<>M!ܭ3M<33>4<EFBFBD>v<EFBFBD>P<EFBFBD><50> <1C>c<EFBFBD>a<EFBFBD><61><EFBFBD>\<5C>3H<33>u(<28><><EFBFBD>0<>X<EFBFBD>N#<23><><EFBFBD>IFÌ1<C38C>Q2<51><32>7<EFBFBD><37>ǯJ 4<>^<5E><>c"<22><>l`w<15>i<EFBFBD><69>\j4W <00><>Ҳ<EFBFBD><D2B2><EFBFBD>q'MI<4D>!4`f<><66><19><><06>}<7D><12><><13><01>)*<02><><EFBFBD><EFBFBD><EFBFBD> <1C>h<EFBFBD><68>㾽1<E3BEBD><31><EFBFBD>þMb<4D>uW <1D><1B>0<EFBFBD>1<EFBFBD>Q<EFBFBD><51><EFBFBD><EFBFBD>bb 8<>E<>F<>c<EFBFBD><63><EFBFBD>m<EFBFBD>C<EFBFBD>m<EFBFBD><6D>Ƌ,<2C><><EFBFBD>(Ӏ<>(<28><10><><EFBFBD>TN<54><4E> éh <0C><06>f
|
||||
;<3B>tda<64>i<EFBFBD><69>1Lj<31>0<EFBFBD>P<>=<3D>1
|
||||
<EFBFBD>8<EFBFBD><EFBFBD>-+x<> K<><4B>"`jGq<47>#<23>z<EFBFBD>[hk4i<><69><EFBFBD><EFBFBD>6<><36><EFBFBD><EFBFBD>Ғ <1C>@ <20>Igy<18><13><>B{2x <1C>1<EFBFBD>4`0 <1E>H<EFBFBD><48>-IwZƘte<74>. t<0E><>Ï<EFBFBD><18>h<EFBFBD><68>f|TА<>(<28>1
|
||||
<EFBFBD>D<EFBFBD>4S<EFBFBD>+o<0F>X2͖"<22>ύWVwf<77>EA<45><41>ü:<3A><>A<EFBFBD>»1 fu<66>B<EFBFBD>0*<2A>Uʟ<55>Nu&C&<26>+<2B><><EFBFBD><EFBFBD>-r<>7 SX<53><58> O<><4F>G` w<>e⇧Fc"<03> <>ZA<01><>e5©<35>8;v<11>$<24>l<EFBFBD>&<26><>,ESkz<6B>b<EFBFBD>/<2F>S<EFBFBD>e<EFBFBD><65><18><>I<EFBFBD><49><EFBFBD>I<01><><EFBFBD>
|
||||
<EFBFBD><EFBFBD>| <08>n<EFBFBD><6E><EFBFBD>PS <1E>"<22>^<5E> <20>f#<23>-<2D>8 <0C><18><>EZI<5A>iƓNɏwUq<55><71><EFBFBD>K<EFBFBD>pݹ<70>';]<5D>(<28>2<02>ܮL<DCAE><4C>NT<4E>ICxrf<72>1)c<>%o<>2~鮺A<S9p#fKdR*<0E><><EFBFBD>xk<78><6B>#<23>ʼn۔ <1F>̑ԥHy" hP <1F>sF<73><46>75<37><35><11><><EFBFBD>&<26><><EFBFBD><EFBFBD>4<>.<2E><><EFBFBD>/<2F>4]=<3D>!<21>R<EFBFBD>(Q!+ <20><>se <0B>o<1B>xs"O<>)9<><39>(<28><>M<EFBFBD>_c<5F> <17>Ԣ<>h$<24>0!<08>= p*4<><0F>0<>h<EFBFBD>G2lQ<6C><51><EFBFBD>s<EFBFBD>x<EFBFBD><78><EFBFBD>bqϞ<> l4<6C>#<23>,1W ;HUhxCy<43><79><EFBFBD><EFBFBD>`<60><>:<3A><06>_ixウ <09>ܩ<EFBFBD>l<EFBFBD>V<><56>gƔ+f|ғ lP<6C>qx<71><78>Z<EFBFBD><5A><EFBFBD>Q<EFBFBD>6?幀<>o$R<><52><EFBFBD><EFBFBD><EFBFBD><EFBFBD>h*<2A><>Ic<49><63>RK(
|
||||
<EFBFBD>)c<>ðp<C3B0><70><EFBFBD>+<2B>ь0 <1E><><EFBFBD>m{<7B>:!<21>ݘQKa=;<3B>[<5B>0<EFBFBD>C"<22><>d<EFBFBD><64><EFBFBD>!<21>
|
||||
<EFBFBD>Y<><59>7&<26>aŀ<61>\ <20><>:Cˉ{劗<15>J<EFBFBD><4A>]<5D>,Tw<54>
|
||||
{X|<15>3<EFBFBD><33><EFBFBD>,<2C>1LA g<>j<><6A><18>º_<C2BA><5F>pL-.<2E>`вX<D0B2><58><EFBFBD>`&<26>sЩ<73><D0A9>>c<>C<1A><>̫<1A>jBk7<6B>[⻏p<E2BB8F><14> u0<75> <00><>)<29><><EFBFBD>ox<6F>#ʍ<>
|
||||
<EFBFBD><EFBFBD>^<5E><> <1E><01><>7<EFBFBD><37>B<EFBFBD>*R#L[<5B>&<26>$<24><><EFBFBD><EFBFBD><EFBFBD>0eAW<41> <1C><>U<EFBFBD><55>쯤Vɤ <16>A<EFBFBD><41>yՓ <1D><>:<3A><><EFBFBD><EFBFBD>5D<35>
|
||||
<EFBFBD>]<5D>JA<11>ޓ<17>)73£<33> E<0E>k<02><>y<EFBFBD>P<EFBFBD><50>9<14>&+<2B>y,l<><6C><15><><EFBFBD>>n <1A>0x<30>*T<>Of]T?]Wnk<6E><6B>ö$t
|
||||
<EFBFBD> z<>2Y<32><59><EFBFBD>{s6^<5E><>U<06><><1B><><EFBFBD><EFBFBD><EFBFBD>d(<28><><EFBFBD>:7
|
||||
B<>C\G"<22>o<EFBFBD><6F>H<EFBFBD>(<28><>-<2D>#8<<3C>c0<63>gt<67><74><08>7<> -,<2C><>I<EFBFBD>_5<02>Jz<4A><7A>rf<72>C<EFBFBD>A<{<7B>4<EFBFBD>2<EFBFBD><32>S8N<38>[<5B>q c,<03><>;@=<3D>N<04>8<EFBFBD>2<EFBFBD> <1D><><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><10>'<16>č<EFBFBD><C48D>'<27>M<EFBFBD>E<07>M<EFBFBD>}c<><63>Y<EFBFBD>'<27><><EFBFBD><EFBFBD><EFBFBD>1<EFBFBD>1<EFBFBD><31>ĺp<C4BA><70>륟<EFBFBD>=<3D>2<18><><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>ad<13>L<0F><><EFBFBD>X<EFBFBD><58>8<EFBFBD>#c,<2C>:8<><EFBFBD>D a<>e<14><>X,<2C><><EFBFBD>(. <20><><EFBFBD><EFBFBD><EFBFBD><EFBFBD>E<16><><15>'<27> <02><> <00>8<EFBFBD><38><EFBFBD>T<EFBFBD>wv*<2A>0' ׅn.F<>s<EFBFBD>0fvúP-gX<><58><EFBFBD>Y<EFBFBD>K&<26>ʮd<CAAE><64> c,<2C>1<EFBFBD><31><EFBFBD>]<5D>89<38>sa,#8<>`t<> <1C><05>#<23><>ƺ<EFBFBD>ɺ<EFBFBD><C9BA>ZE<5A><45><EFBFBD><EFBFBD>M<EFBFBD><4D><EFBFBD>:<3A><>=D<>N#n5<6E><35>jbW<62><57><EFBFBD> <1E><><EFBFBD><EFBFBD><EFBFBD><EFBFBD>EV <04><>?<3F>q<EFBFBD><71>9<EFBFBD>:Y<><59>[<5B><><EFBFBD>:G<>
|
||||
\<5C>x<EFBFBD>K<><4B><EFBFBD>P<EFBFBD>- )P<>%P<>T6<54>z<EFBFBD>j<EFBFBD><6A><EFBFBD>h<EFBFBD><68>Ȏq<C88E>"
|
||||
<07>Ri0<69>q<EFBFBD>$g<><67>5<06>s<EFBFBD><73><EFBFBD><EFBFBD>}g]<5D>J<05><><EFBFBD><EFBFBD>@<40>`<60><><EFBFBD>ຨ$<24>:<3A>516<31>ĵE'<27><><EFBFBD>m<EFBFBD><11><><EFBFBD>SU<53><55>0,<2C><> s<>Ab<>reߍUvP<76><50>$`<60><11>Z<EFBFBD><5A><EFBFBD><EFBFBD><EFBFBD><EFBFBD><11>/<2F>
|
||||
WB<EFBFBD>X <0C>[<07>/F̘<>͵<EFBFBD><CDB5><EFBFBD><EFBFBD><EFBFBD>nC8w<38>-ŚE*ɾ<><C9BE>[)<29><>7<EFBFBD>H<EFBFBD>~<7E>]<5D>dLCi<43>x
|
||||
<EFBFBD><EFBFBD><EFBFBD>Q<EFBFBD><10><><EFBFBD>1<EFBFBD><31><EFBFBD><18><><EFBFBD><EFBFBD>1<EFBFBD>4<1A>5<EFBFBD>G$&<26><><EFBFBD><<3C><>h<EFBFBD><68> <09><18> <09><>^<5E><><EFBFBD><EFBFBD>/G\<5C><><01><><EFBFBD>EVh<56><68>?<3F><>J<>Iu<49>M<>dN<>3w<33><77>Z<10><>9]ߕ<><DF95> =<3D> <0C>A<EFBFBD>A<EFBFBD>A <><7F> 5<><35>,<2C><>td <1E>y<EFBFBD><79><EFBFBD><07><>3 f~<7E> <19>-<1A>ª<EFBFBD>L<EFBFBD><18><14><><EFBFBD>`<60>+$?C<>gg
|
||||
<01><>W<EFBFBD>!F4<46><34> %nLsnj<73><C78C><EFBFBD><EFBFBD>T|<7C>B,<2C>@<40><>z\ <0C><>j<EFBFBD>z+7-A<><41><EFBFBD><EFBFBD><EFBFBD>:X(<28><><EFBFBD><EFBFBD><EFBFBD>A<0E>|2<>Ѱ P<><50><EFBFBD><EFBFBD>Ѓk<><6B>X<EFBFBD>Z<EFBFBD>w<EFBFBD>]<5D><><EFBFBD> 3"<22>Ŝ<EFBFBD>F7<46><37><14>I<EFBFBD><49><EFBFBD>{<7B>3<EFBFBD><33><EFBFBD>Z(<28>,<2C>F<EFBFBD><46><EFBFBD>q{EU<45> )\ <1D><<3C><>k<><6B>πۄ<CF80>H<EFBFBD>^<5E><><EFBFBD><EFBFBD>
|
||||
<EFBFBD><04><>+}<1B><><EFBFBD>"<22><><EFBFBD>%M<>G<EFBFBD>:<3A><> <1C><><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>b<>l9<>!<21><0F>S<EFBFBD> t<><1A>6<EFBFBD><36>Ψ=
|
||||
:c<><63><EFBFBD><15>Ƥ<>r<EFBFBD>Y[<03><>-p&yX1<08><>)<29>A<EFBFBD><41>p<EFBFBD><70>3<EFBFBD><06>_<EFBFBD><5F><EFBFBD>d<16>+s<16><06><><EFBFBD><EFBFBD>`<60>R<EFBFBD>k$ұ<>@<40><><EFBFBD>{zdMJ<4D><4A><EFBFBD><EFBFBD>*S<>Lpv4k<34><1A><><14>*L<><4C><EFBFBD><EFBFBD>r&<26><>(<28> TA _(\<5C>Q<EFBFBD>I<EFBFBD><49>4<EFBFBD><34><EFBFBD>ps <1D><><EFBFBD>a<EFBFBD><61>g1<><31>NP<4E><50>W<EFBFBD>5<EFBFBD>g<EFBFBD><67><05><12>oa; t<><74>3t!m$ <0B>N2<15><>~zV
|
||||
<EFBFBD>7m<>ٜE<><03><>(Ѱ<>Y#<08>8+<2B><> <1D>p<>bӣpC <1C>o?3l<33>i0i<30><69>L<EFBFBD><4C>Ԫ<EFBFBD>Ue<55><65><EFBFBD>g8<67> H#<23><>c=,<2C><>͏%]<5D>4<><34>5.=$<16><>qf<71>}<7D><>Nt<4E>8<EFBFBD>G <12><><EFBFBD><EFBFBD> <1E><><EFBFBD><EFBFBD><EFBFBD><EFBFBD>Z'<27>KYq7<71><37>W<EFBFBD><57>l % <20>5F<35>u <20>#<23><><EFBFBD>M<EFBFBD><4D>I+y(<28><><EFBFBD><EFBFBD>~,<2C>h<EFBFBD>b<14><><EFBFBD>s9<73>
|
||||
`B_%<25><><EFBFBD>2<EFBFBD><32>cp<63><<3C><><EFBFBD><EFBFBD>
|
||||
!<21><>"<22><>jK%><3E><*<2A>`<60>;<06>l<EFBFBD>c<EFBFBD><63>@<40><><EFBFBD>2S<32>]\<5C>SiA<69><EFBFBD> <1E><>Sփ$ <12>!e'<27>V<EFBFBD><56>`sB<73><42><EFBFBD>T<EFBFBD>
|
||||
<EFBFBD> <20>36.D<06>J<EFBFBD><4A>N<4E>)<29><><EFBFBD><EFBFBD><EFBFBD><2<> H9<48><39>B<EFBFBD>B7<42><37>y<EFBFBD><79>@<40><>9<19><>lb<11>c<EFBFBD><63><EFBFBD><EFBFBD>9<EFBFBD><39><EFBFBD>2<EFBFBD><32><EFBFBD><EFBFBD>(<28>(<28>酶Y<>R<><52>3<EFBFBD><33><EFBFBD><EFBFBD>-ј<04>2<EFBFBD><1A>N<EFBFBD>0X<30><58><EFBFBD>pF<<3C>h<EFBFBD><68>:ÚN{|<7C><>[<5B>R<EFBFBD><52>Zs2OUB<55>,V<>
|
||||
,<14>l <0C><><19>V<EFBFBD>S<EFBFBD>=<3D>p C<43>v<EFBFBD>~<7E><>4<EFBFBD>1<EFBFBD><31>FefO:
|
||||
L<EFBFBD>/<2F>bB<62>y<EFBFBD>S^<5E>Jc:<3A><><08><>3<EFBFBD>A<>-<2D>.K<>k$,<2C> oBj<42>v<EFBFBD>c?1<>O<><4F><EFBFBD>JKbGv/<2F>'<27><><EFBFBD><0F>Ґf d2!, <1F> <20>!
|
||||
iy<17>Fu<46>D<EFBFBD>f"<22><><EFBFBD>j#Yˎ<59> b\<02>6
|
||||
k@<40><><1A><> <19><>
|
||||
<EFBFBD><!V<>Z<EFBFBD>"<22>'<27>ȼKN<4B>Q<EFBFBD>J<EFBFBD> %
|
||||
<EFBFBD><EFBFBD>
|
||||
!<21>0(<28>z<EFBFBD>ʘ<04>Ζ5ꤍ<35><EAA48D>)!<21><>t<EFBFBD>iTf<54>x<EFBFBD>A<EFBFBD>*<2A><><EFBFBD><EFBFBD>#dkJ<6B>L<EFBFBD><18>ThM<68><4D><EFBFBD>G<EFBFBD>c<EFBFBD>f<EFBFBD>;<3B><> <20><>+xUݟ<1B>*o<13><>;o<0F>ne0<65><30><EFBFBD><EFBFBD>M<EFBFBD>`<60> <00><>Z<EFBFBD>6<EFBFBD><36>P<EFBFBD>FA<46><41>4J <10>,<2C><> U^T<>+5:<3A> y<><79>`<1A>sa<73>
|
||||
<EFBFBD>b"1WA<10><><EFBFBD>\Bp<>(n<05><>*<2A><<3C>o<EFBFBD>M<EFBFBD>,<2C><>A+<2B>
|
||||
}fA<10>e<EFBFBD><65><EFBFBD> C(<28>=<05>aR<>1
|
||||
A<EFBFBD><EFBFBD><EFBFBD>(H<><48><<3C>J<EFBFBD><4A><EFBFBD>O<EFBFBD>QB,Fe<46><65> V<>Ev%!j^<5E><>=c<><07>E$<24>i<EFBFBD><69>5[<5B><><EFBFBD>A<EFBFBD>@$_AM<41><4D>T\<<3C><><EFBFBD><EFBFBD><EFBFBD><EFBFBD>QP`<60><><EFBFBD>Jޜ<4A>%D<>ZQ "P F<><18><><07>z5<7A> <1E>x#@<40><>Ļ<EFBFBD>*<1B> <1E><>2<>)<29><>z<10>L<EFBFBD>d<EFBFBD>/fy<66><79>"<22><>kqq<71><71><EFBFBD>7U3<55><33><EFBFBD>(L<>e <1D>\#ƻ<>$u<>-\<17> _<><5F>f*K<><4B><EFBFBD> <1E><><EFBFBD>@<40>F`<60>:1<><08>*<2A>"Ǡ<>h<EFBFBD><68><EFBFBD><EFBFBD>^+<2B><>]n<>l1No<4E>C<EFBFBD>%&(<28>><3E>pz;<3B><>u <09><><EFBFBD>mS<6D>Y+<18>Z<EFBFBD>p<><70>g<EFBFBD>]/<2F>V<><56>x1<78><31>H<><48>`<60><>%ag<61><67><EFBFBD><EFBFBD>,<2C><><EFBFBD><EFBFBD>[<10>
|
||||
<EFBFBD><EFBFBD>D<EFBFBD>\ʕ<><CA95>8<>)<29>h<EFBFBD><68><EFBFBD>ą<EFBFBD><C485><02><>
|
||||
<EFBFBD><EFBFBD>[8[ 0<>Gc
|
||||
<EFBFBD>.<2E>X<EFBFBD>7<EFBFBD><37><0F><>P<03><><EFBFBD>B<><42><EFBFBD>r^<5E><><EFBFBD><EFBFBD><EFBFBD>h.<2E><>!c<>F}QtB@g<>wM/sm>D<><44>AU<41>Ms<4D>D<EFBFBD>J ̜ɚ<CC9C>3<EFBFBD>
|
||||
iH<EFBFBD><EFBFBD> {Rj!S<>=M<>&<26><>>X<><58>!*ٺ@c
|
||||
<EFBFBD><EFBFBD><EFBFBD>F<>Ex*<2A>
|
||||
<EFBFBD><EFBFBD>X<EFBFBD>&<26><><EFBFBD>Z<EFBFBD>X<>E<>-!,<2C>n\A<>50<35>!<21>S<EFBFBD><53>l<><18><>)<1B>G<EFBFBD><47>I<04><><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><0E><14><>P<>Y<EFBFBD>p<EFBFBD>ߜT<DF9C><<3C>uT;<3B><>f<EFBFBD>2<><32>~6<>4u^<18> <13>իL<D5AB>C
|
||||
M<EFBFBD>a\<5C>{<7B>%f<01>m<EFBFBD><6D><EFBFBD><EFBFBD><EFBFBD>B<7F>ʨ%Ƈ><3E><><EFBFBD>ԭƬ<D4AD>$ Z,<2C>D<EFBFBD>X<EFBFBD><58><EFBFBD>
|
||||
<EFBFBD><EFBFBD>:riǐ <0B><>
|
||||
H<EFBFBD><15><><EFBFBD><EFBFBD>lͿ<6C>x<EFBFBD>t<1B><>! <1D>4<EFBFBD><34><EFBFBD>Y$M<>\+<2B><><EFBFBD><EFBFBD>1<EFBFBD>P,<2C><>)d<><02>4<EFBFBD> <20><>Mj<4D><6A>i 究;<3B>
|
||||
<EFBFBD><EFBFBD>q3<EFBFBD>2><3E>S<EFBFBD><53>QC/<2F>
|
||||
S!8<><38> <1C><>d!<><7F><EFBFBD>w<EFBFBD>u<EFBFBD><75>0^<5E><><EFBFBD>8*l9<6C>b<EFBFBD><07>"
|
||||
<EFBFBD>4<EFBFBD><EFBFBD>\O<><4F><EFBFBD><EFBFBD> <1C>'"<22><07>T <1D><><EFBFBD><EFBFBD>X<EFBFBD><58>C<EFBFBD>1<EFBFBD><31><EFBFBD>k<EFBFBD><6B>q<EFBFBD><71><EFBFBD><EFBFBD>[<5B><>NX<4E><58><13>pf&<26><11> <0B><><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD> )5<><1A>;<3B><>w<EFBFBD>r<>4^<5E>w<EFBFBD>n<EFBFBD><6E><EFBFBD>\+<2B>c <0B><>]<5D> <1E><> <1E><><03>u<>qv4<76><34><EFBFBD>L<EFBFBD>@3<>Q)<29>j<EFBFBD>0<EFBFBD>y"<22>q<07>A<EFBFBD>ۓ<EFBFBD><DB93><EFBFBD>6<EFBFBD>a<EFBFBD><61>\'<27><><EFBFBD>yH $0<><30>#<23>4<EFBFBD><34><EFBFBD>V<EFBFBD><56><EFBFBD><EFBFBD>Lu<<3C>i<EFBFBD><69>7<EFBFBD>s<><73><EFBFBD><EFBFBD>a<EFBFBD>nA)<29>-1<><31>':@<40>a<EFBFBD>r<>h<EFBFBD>(<28>b<EFBFBD><62>#<23>lf1<66>{<7B>8/f<>1<EFBFBD><31>;<3B><>
|
||||
<EFBFBD>9S`<60>k<EFBFBD><6B>0X$<24><><EFBFBD>s 9<><39><EFBFBD>[<5B>苲~<7E><><EFBFBD>NoF<6F>5Ʊ<35><C6B1>3;<3B><>q<EFBFBD>,<2C>.<2E>1z<31>XBWb s<>-<03> <0C>8Y<38>bl<62><6C><EFBFBD>U<EFBFBD> 9(<28><[ <1C>uLRX<52><58><EFBFBD>c<EFBFBD><17><><EFBFBD>N<EFBFBD><4E><EFBFBD><EFBFBD><EFBFBD>m<05>A<EFBFBD><41><06><01>]<5D><><EFBFBD>a<EFBFBD><61>$I}b<><62><EFBFBD>f<EFBFBD>8<EFBFBD>+<2B>uF<75><46>BU?h <0C>Lsġ1<C4A1><31><EFBFBD><EFBFBD><EFBFBD><EFBFBD>x"<22><><EFBFBD><EFBFBD> v<><76><EFBFBD>b<>;<3B>=9<>~<7E>5<16><CH3<48><33><EFBFBD>fn3<6E>6<EFBFBD><36><EFBFBD> ;tB<74>ɋˤk<CBA4>[:?s<><73><EFBFBD>q,<1B>j<EFBFBD>F<EFBFBD><46><EFBFBD><EFBFBD><EFBFBD><EFBFBD>b<EFBFBD>>C[<5B>i<EFBFBD><69>N<EFBFBD>+<2B>mf<6D><66><EFBFBD><EFBFBD>`F<>h <1E>"+<10>2 <20>҈+(1<><31>J<EFBFBD>v:c<>
|
||||
^<5E>18<31> 415<><07><><EFBFBD>2z<>iF<16>b<62>(<28>j,<12>=<3D><><EFBFBD>QEv7<>ć<EFBFBD>΄<EFBFBD><CE84>h<EFBFBD>ƂhN<68><4E>=<3D><>n(<28>ҋ}<7D>Br<42><72>h,k<>yף $ a<><61> <0B>LR<4C>cF<63><46>) S<>,<2C>k ;<3B><>W<EFBFBD>1<EFBFBD>@<40><><EFBFBD>a<EFBFBD>ѹ<EFBFBD><D1B9>̬<EFBFBD>/1<><31><EFBFBD><EFBFBD><10>
|
||||
L̙1 OS<4F><53>oE=1<>3<EFBFBD><33><EFBFBD><EFBFBD><EFBFBD>,m<><6D><EFBFBD><EFBFBD>-Hׯc%x<><78>!lB<6C><42> <04>睛F<E79D9B><46>IM<18><>ke><3E>]<5D> G <1F><><EFBFBD><19>.
|
||||
<EFBFBD><13><><EFBFBD><11>XChX<68><58><EFBFBD><EFBFBD><EFBFBD>aD<61><44>7<EFBFBD><37><EFBFBD><EFBFBD>nbC<62> f<><66> <0C>ώ<EFBFBD>0<EFBFBD><30><EFBFBD>a%<25><><EFBFBD><EFBFBD><EFBFBD>hSө҆G<D286>]<5D>l<EFBFBD>5<1A>j<EFBFBD>t<EFBFBD><74><EFBFBD><EFBFBD>yb<79>M<EFBFBD><4D>xo<78><6F><EFBFBD><EFBFBD>>c<18> <0C><>r
|
||||
<EFBFBD>}<0F><>6<EFBFBD><36>0<EFBFBD>R <08><><EFBFBD>,nB<6E>Y<EFBFBD> <11><>Ci<43>z6<7A>Շ{<7B>^E= <00>j<EFBFBD><6A><EFBFBD><EFBFBD>c3<63><33>WJ<57><1A><1B><>d<7F>-k\<5C><><EFBFBD>|<04><>W<EFBFBD><57>re<72>۾o 64I<34><49><10>W<EFBFBD>a %9Q<39>U<15>2Ɛ<32><C690> <1E>Y<EFBFBD><10>t/H<>7<EFBFBD><37>,ד<>J<EFBFBD>j<EFBFBD>*<12>軳<EFBFBD><E8BBB3>*<2A>m <0C>:M<><4D>-~/<2F><><14><>[<5B><08><>³<EFBFBD><C2B3><EFBFBD>^c<><63>!<21>&RIvB8<42><38>Y <0C>_fI<66><17><>,i<><69>0b-)<29>~<7E><>C<EFBFBD><13><><EFBFBD><EFBFBD>*<2A>F<>٩~<7E><>\S|%<25><><19><>4<EFBFBD><34>l<EFBFBD>v*)<29>=c<><63>I<05><EFBFBD>@M
|
||||
<EFBFBD><EFBFBD><19><><EFBFBD>&<26> >W <0C>_9߆*C<><43>
|
||||
Cj!p
|
||||
<EFBFBD><EFBFBD><EFBFBD>c<EFBFBD>^<5E>a<EFBFBD>V<EFBFBD><56>l :00?c<18><><EFBFBD>K<EFBFBD>1<EFBFBD> <00><><02>pƊ1<C68A> Y]<5D>5<0E>Z<EFBFBD><5A>k<EFBFBD><6B>? <1C>Ws 5 <20>T<EFBFBD><54>0O<30>́1<CD81>c<1A><>5tz͓ <1E><10>%<25><>봚l<EBB49A>k<EFBFBD>u<EFBFBD> <1E><B<><42><EFBFBD>H<EFBFBD>1<EFBFBD><31><EFBFBD>AT/c<><63>ZR@Ь|'1M<31><4D>I<EFBFBD><49>~g<><67>P`<60><18><>9j<39>%<25><><EFBFBD>ʟ<EFBFBD><CA9F><EFBFBD><EFBFBD><EFBFBD><EFBFBD>S<>q
|
||||
<05>O%w\<5C>%Zg<5A><67>3!<21>E<EFBFBD><45>0<EFBFBD> کf4<66>h}<7D>/Ԙ<><D498>N[g<>8f<38>C<EFBFBD>n<EFBFBD><6E><EFBFBD>*<2A>-!<21>T=<3D><>NA<4E><41><EFBFBD><13>*<2A>5,La_<61>F;<3B>/٨<>~|B<><42><EFBFBD>
|
||||
X<EFBFBD><EFBFBD>T<>Tq<54>A<EFBFBD> <09>t64A<34>Q<EFBFBD><51><EFBFBD>G<EFBFBD>\<5C><0E><11><>vĈq<C488><71><EFBFBD><EFBFBD>qt<71>&a<>8<EFBFBD><38><EFBFBD><EFBFBD><EFBFBD>1a_<61>1<EFBFBD>r<EFBFBD><72>GW<47><57>FT<46>^V<>1<15><><EFBFBD>mY#<23>]C<11>TG<54>O6a<><61> <20>"<22>:<3A><1A><><EFBFBD>9<EFBFBD><39><EFBFBD><EFBFBD><EFBFBD><EFBFBD>+<2B>Nc<4E>p<EFBFBD>]<18><11><>k<EFBFBD><6B>ȁ<17>pt'E<><45><16><><EFBFBD>><3E><><EFBFBD><08><><EFBFBD>W<><H<><48> '[*<2A><18> 7<><37>e2<65><32><EFBFBD><04>=<3D>?><0E>m<>1 <09>F<EFBFBD><46>=<3D><><EFBFBD><EFBFBD>:C<17><>X<EFBFBD><58><EFBFBD>o <1C><> <12>n<EFBFBD>><3E>S<EFBFBD>
|
||||
<EFBFBD><EFBFBD>c<EFBFBD>h2&<26>W<EFBFBD><57>Qn<51><6E><0E><><EFBFBD>4<EFBFBD><10><><EFBFBD>c6<><36><EFBFBD><EFBFBD><EFBFBD>Ô<EFBFBD><C394><EFBFBD>Sw<53>"<22>Ę<EFBFBD> <1E>y2.<2E><≯}X <09><><EFBFBD><EFBFBD>c`<60><><EFBFBD>=<3D>Mwa<12><>D<EFBFBD>u(<28><15><>
|
||||
&ش<>䉰<EFBFBD><E489B0><EFBFBD><EFBFBD>N<EFBFBD>;<3B><>H<EFBFBD>b<EFBFBD><62>;6t<36><74>LcH5<48>%:O<>;<1B><><EFBFBD><EFBFBD>
|
||||
<EFBFBD><EFBFBD><EFBFBD>~<7E>TUsSa<53><61>(<28>!<21><><EFBFBD><EFBFBD><EFBFBD> <20><0F><><EFBFBD>T<EFBFBD>z<EFBFBD><7A><EFBFBD><<3C><>|<7C>N<EFBFBD><4E>P<15>5<17><>d<EFBFBD>P I<><49><EFBFBD>r<EFBFBD><72><EFBFBD>v<EFBFBD>-<2D><><EFBFBD>!cNP<4E><50><EFBFBD>y^ <1E>N<EFBFBD><4E><EFBFBD><EFBFBD>3<EFBFBD>o^<5E>:<3A>"<22>r<><72><EFBFBD>z<EFBFBD>D<EFBFBD>p`<60>Z<>8<EFBFBD>L<EFBFBD><4C><EFBFBD>0<05><>9<EFBFBD>qv<> V<>M<EFBFBD><4D>kأa$1<><31><EFBFBD>c<06>؎.<2E>NH<DEB2> <09>U<EFBFBD>r;<3B>D<EFBFBD> <1F><>q<EFBFBD>th<>Za9<61>j<EFBFBD> ]I<><49><EFBFBD><EFBFBD>A<EFBFBD>_<EFBFBD>` twO<77><4F><EFBFBD>H<EFBFBD>2<EFBFBD>8<EFBFBD>g<EFBFBD>@<05><><EFBFBD>=<3D> ӏ<><D38F><EFBFBD><EFBFBD>JU<4A><55><EFBFBD>_<EFBFBD><5F><EFBFBD>X<EFBFBD><58>T <1F>p;7O<37><4F><EFBFBD>̒<EFBFBD>Ǹ.<2E><>ذ
|
||||
<EFBFBD> S<>q<12>p<EFBFBD><70><EFBFBD><EFBFBD><11><>/H<02><><EFBFBD>Gx=8s f`<60><>3<>0<EFBFBD>?<3F><><EFBFBD>:<3A><>;"<22><>J|SҘp<D298><70><11><>qWBb<42>Y<EFBFBD>Y<EFBFBD><59>R<>5c1;[<5B>^<1B><>. <1C>5GW<02>N<EFBFBD>"O<> <1E><>2<EFBFBD><32>e<EFBFBD>5<EFBFBD><35><EFBFBD>G <0B>MG<4D>9<EFBFBD><39>I<EFBFBD>Cfdg<64>T*6 <0B>mG<6D><47><EFBFBD><EFBFBD>4<><34>c\<5C><><EFBFBD><EFBFBD><EFBFBD>XBD<42><44>S<><53>% ){oa<6F> qZv<5A><76>f-ɸD-<2D><><EFBFBD>X3<58><33><EFBFBD>J{`<60>]s <09><08>j7ƪk+<2B>XI<58>aָ^Y<>]M<02>鐢
|
||||
;<3B><><EFBFBD>t '<27>u{B<>Wǹ<57>t<0F><><EFBFBD>s'Y<>z`C]<5D><>I<EFBFBD>a<EFBFBD><61><EFBFBD><EFBFBD><EFBFBD>H<EFBFBD><48><EFBFBD>o@<18>a<EFBFBD><61><EFBFBD>t<EFBFBD>_Y<5F><59>-Ŏ<><C58E>P}o<>, ys<79><73><EFBFBD><01><>9La0븰~<7E>wf<77> <1D><><18><><EFBFBD>9b<39>K<EFBFBD><<3C><> X<><58><EFBFBD>8z<38><7A>zal4[{(wxüPudY<64>S<EFBFBD>1!<21><> <<3C><><EFBFBD>t)w<><77><11>'<27><>X<EFBFBD><58>Uc0|<7C><>P<18>Pn<>ˌ<EFBFBD>ȭ<EFBFBD> <0B><05>=<3D>c TD<54> <0B>=<3D>X<EFBFBD><58>XH<><48>X'<18>`g<>qc<71>S>}<7D>&<26>&Gꄩ'M̓I<CC93>lO1c<31>K<EFBFBD>/<2F>Mv<15><16>8<EFBFBD>v9<><39>@<40>F<>&<26><><EFBFBD><EFBFBD>X<EFBFBD> rDH<44><48>~`.<2E><> <1E>y<>tf <0B><19>MM <1E>ث<EFBFBD>uƽ0<C6BD>W<>0<EFBFBD><30><EFBFBD><EFBFBD>+<><DB9D><18>p<EFBFBD><70><EFBFBD><11>u<EFBFBD>[<5B>q<EFBFBD>^qYc<59><63>h<EFBFBD>U@<40><><EFBFBD>Z1<16><05><>1<08><><EFBFBD>%$<24>f<EFBFBD><66>ƏU<C68F><55>j<EFBFBD><6A>:)<16><><EFBFBD>K<EFBFBD>Te<>c<<3C>=.A3<41><33><EFBFBD>r<>`<60>ax<61><78><EFBFBD><EFBFBD>i'<27>| <1A><><EFBFBD>mD<6D><44><EFBFBD>6<EFBFBD>҅u5<75><11>pFg<19><>X<EFBFBD><58> <18>uτ<75>7<><37><0F>p<EFBFBD>ڡx <1C>O#<23><>i~ <1F><>G`L+={`tPOc|Tch<63>i<69>3] <0C>߅<EFBFBD><DF85><EFBFBD><EFBFBD><EFBFBD>{\<5C><><EFBFBD>H<EFBFBD>+1}<7D>R<EFBFBD>[3<>u<><75>7<EFBFBD>3<EFBFBD><33>5<EFBFBD>8諒<38><E8AB92>hJ53R<18>w<EFBFBD>[<5B>դ<EFBFBD>f<EFBFBD>-<2D><><EFBFBD><EFBFBD>0<EFBFBD>3Ne&<04><>$<24><><EFBFBD>h<06>z=<3D>a<EFBFBD>yς<79>7%)n P<>"}a*<2A>)<29><1A><>P<EFBFBD><50><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD> b<>=<<3C><> <0C>5<EFBFBD>q<EFBFBD>}ol<>,<2C><>8<EFBFBD>N%V<><56> 5<>9c.<2E><><EFBFBD><EFBFBD>`ƙջИ<D5BB><D098><02><>X<EFBFBD>]<5D>ܞ\<18>H3<48><33><EFBFBD><EFBFBD>դ<EFBFBD><D5A4><EFBFBD><18>p<EFBFBD><70><EFBFBD><EFBFBD><EFBFBD><EFBFBD>z<10>wV*<2A><>t1Y-"<10><>it<14>/<2F>^<5E><>F8<46>5<EFBFBD>y<19><><16>A <20><14>D<EFBFBD>F<EFBFBD><0E><><EFBFBD>%*\<5C>KL<4B>cS<63><53><17>Y<EFBFBD>e<EFBFBD>{-rfH&<06>5<EFBFBD><35>/G<>a f<>՞q$<24>-
|
||||
<EFBFBD><EFBFBD><EFBFBD><18>h<EFBFBD><68><EFBFBD><EFBFBD> O<>3c-<2D><><EFBFBD>X<EFBFBD><58>\<5C><>Ƙ<EFBFBD>3<EFBFBD>y<EFBFBD><79>G<EFBFBD><47>>f)*<2A><>]<5D><><EFBFBD><11>g5<67><35>ƹ?0<><14>}4<><34><EFBFBD><EFBFBD>q'<27><19><>z<EFBFBD>,<18><>_u<5F>]1<1B>L1<4C><01><>,<2C>PV<50>S<EFBFBD><53>a38/<2F>+<1B>4u<34><75><EFBFBD>
|
||||
<EFBFBD><EFBFBD>ea& <1F><><EFBFBD><EFBFBD>de*<2A>Vr<56><72>Ȇ8y<38>K<02><><0E>y]<5D> 1L{<7B><><EFBFBD>Τa<CEA4><61> <1C>0<EFBFBD><30><EFBFBD>{,<2C>R<0F><>h<EFBFBD>nW{;c8tx<74><78><EFBFBD>g<EFBFBD><67><EFBFBD><EFBFBD>C/<2F>G<EFBFBD><47><EFBFBD>a<EFBFBD>5<EFBFBD><35><EFBFBD>d0<64><D4<44><34><EFBFBD>TLB<4C>^ԕ6d<36><64><06>P <00><>E<EFBFBD>0<EFBFBD>h<EFBFBD>GHX'5<><10><><EFBFBD><EFBFBD>/<2F><><EFBFBD><12>
|
||||
JE<><45>k<EFBFBD><1A>c,Í7
|
||||
<EFBFBD><EFBFBD><EFBFBD> <1F>ۻ$<24><>}Y<><59>yɁQ=n<><6E> ^<5E>wN<77><4E><EFBFBD>h a<><61>P<EFBFBD>2<EFBFBD>b<EFBFBD><62>6qǰ%<25><>|<7C><><EFBFBD>7FN<46><4E>
|
||||
cR@b ^<5E><><EFBFBD>f<EFBFBD>`,<2C>K<EFBFBD>^Wa<57>$EhuԔ<75><D494>$0l"W <10><><EFBFBD><EFBFBD>bt{8<>
|
||||
<EFBFBD><EFBFBD>:<3A><>l<EFBFBD><6C>z<EFBFBD>]<5D>8<EFBFBD>9<17><12> <1D><><18><1B>nl<6E>3<EFBFBD>nc0<63><30><EFBFBD><EFBFBD><EFBFBD>ڕ<EFBFBD><DA95><EFBFBD><EFBFBD>[X/:<3A>SIC<49>1<EFBFBD>q<EFBFBD>t
|
||||
VƀF<C680> 5<>4<EFBFBD><34><EFBFBD>0<EFBFBD><30>D<EFBFBD>P<EFBFBD>s<>]<5D>?<0E><> Gy<47>#W b<>
|
||||
8.<15>rwc0<63><30><EFBFBD>z+*<2A>%F <0B>-Џq<D08F><71>=:<13><>HGB`<60>R^E ~u<> <0B><>
|
||||
<EFBFBD><EFBFBD>Tנn<EFBFBD>ֶ\<5C><><EFBFBD>q<EFBFBD>(<28>0 <00><>cx <0C>nY<6E><59> <0C>p-<2D> <1C>} <1E>bT<62>յr\<5C>[<5B><>e<EFBFBD><65><17><>Tc
|
||||
<EFBFBD>`<60>[{<7B><1A>0f<30><66>v<EFBFBD>1C<07>9MC<4D>p<EFBFBD>]<5D><><EFBFBD><EFBFBD><EFBFBD>5X<12><>j,<2C><16><><1B>X`5:<3A>lTc<54><63><EFBFBD>V <1B>X`<60><><EFBFBD><07><><EFBFBD>R<EFBFBD>k<EFBFBD>Tc<17>x<EFBFBD>m<EFBFBD><13>%<18>,;+<2B>X`<60>[5ި<35>c)<29><><EFBFBD>i, <20>uf<75>Z<EFBFBD><5A><02><><EFBFBD><EFBFBD>=1<>&~<7E>'<27><><EFBFBD>x<EFBFBD><78><EFBFBD> <0B>X`i<><69>c# <0B>%X<><58>D4f<34><66><EFBFBD>i<EFBFBD>΅h,<2C>q<EFBFBD>9a# <0B>j<EFBFBD><6A>xZ<10><05>k<EFBFBD><6B><EFBFBD><13>X`%*<2A>7<EFBFBD><37><EFBFBD><EFBFBD><EFBFBD>aݍh,0I<30><49><13>X`2nODc<44><63><EFBFBD>0y<30>B4X<><58>=x<>j<>e<EFBFBD><11>V<><15>F4^<5E><><EFBFBD>h,0<05><>x<EFBFBD>:D<><44><EFBFBD>3f <0C><><EFBFBD>6<><36><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD> {<7B><>V<15><EFBFBD> <0B>[<5B><>] <0B>X`9;U|<7C>
|
||||
Jp<EFBFBD>-<c<><63> <0C><><EFBFBD>g,<2C><><EFBFBD><EFBFBD><13>X`.<2E>z<EFBFBD> <0B><><EFBFBD><EFBFBD><EFBFBD><EFBFBD>3M<><16><><EFBFBD><EFBFBD>h^<5E><11><19><>q<EFBFBD>эh,0*j<><16><><EFBFBD><EFBFBD>E<EFBFBD><45><12>v2Nx{" ,<2C><15><11>; <1F>{+<2B>X`<60>/=oODc<44><63><EFBFBD><EFBFBD>~Y<><59><02>bs<1B>X`
|
||||
6ݞ<EFBFBD>ƌ<EFBFBD>#<23><><EFBFBD><16><><EFBFBD><EFBFBD>;<3B>h, <20>%<25>'<27><><EFBFBD><EFBFBD><<3C><>m! <0B>m~<7E><>ƌ<EFBFBD>Ѽ*+<2B>X@-<2D><>n<c<><63> <1D><13>X`<60><>x<EFBFBD>3F<33><0E><><EFBFBD>3I<><49><EFBFBD> <0B>X`r<>ߞ<EFBFBD><DF9E><02>)<29><>V<EFBFBD>1<EFBFBD>]<5D><><EFBFBD>i°<69>h,0g}><11><05>^ҷ'<27><><EFBFBD>Xh<58>oODc<44><63><15><>D4<16>U<EFBFBD><55><11>v5S݈<53>sc<73><63>f̈<66>,ܞh<DC9E>s%<25><13>X`yT?<3F><>3<16><>V<EFBFBD><56>'<27>1c<31>`<60>3q<33> <0B>(A<><41>h,<2C><14><><EFBFBD>3f<33>ٙ<EFBFBD><D999> <0B>X`U<><55><13>X`):m<c<>ԝӏ<D49D><D38F><EFBFBD><05>Ҁ<EFBFBD><13>X`<60>rf<72>ZxƌQ<C68C><51><EFBFBD>n<EFBFBD> <0B>*<2A><><EFBFBD>g,<2C><><EFBFBD><EFBFBD>'<27>1c<31>O<EFBFBD><4F><EFBFBD>h,0<16>| ,<2C> <20>Zx<5A> 1
|
||||
<EFBFBD>;b<> <0B>$<07>W<EFBFBD><57><EFBFBD><EFBFBD>n<EFBFBD><1B>X<EFBFBD>.=<3D>Q<EFBFBD>y<EFBFBD>ae<61>l<7F><13><><EFBFBD>8p<38> <0B>D<07><>g,0<> <0C>:k<>36ؑ<36><D891>g, <20><>ݞ<EFBFBD>ƌuy=<13><05><><EFBFBD>]<5D><>;.<2E><>w<EFBFBD>1#d6ԫYi<59> _<>N3f<33><66>D<ь<05><> <0B><>ӌ<19><>t<06><>;<3B>X@<0F>ەf<DB95><66>9"<22>x<EFBFBD> Hw<48><77><EFBFBD>f<EFBFBD><66>#x<>Ҍ<05><><EFBFBD>d +͘1<CD98><31><EFBFBD><EFBFBD>ӌDL<>~-4c<34>u4
|
||||
<EFBFBD>A3ؙl<D899>n<c<>f"<22>Q<EFBFBD><51>3Nθ<4E><CEB8> <cƆ<63><C686>{<7B><><05><><EFBFBD><EFBFBD><EFBFBD>3f<33><EFBFBD>d<EFBFBD> L<<3C><>=xƌ=<02>Ϙ1<CF98><EFBFBD>Wl! ,SJ<04> 3<>{)<29>f% <0B><><EFBFBD>؈ƌ<D888>y<EFBFBD> ۈƌ<DB88><C68C>qg<71><67>h,<2C>tq<74>-<c<><63> <10><><EFBFBD>#<23><>iGV<47><56><EFBFBD><EFBFBD>!Y˲<59><CBB2><11><><EFBFBD><17><><EFBFBD>ܞ=<3D><07>X`Wv<57><76>3f<33><66><EFBFBD><EFBFBD><EFBFBD><EFBFBD>3f<33>M<EFBFBD><4D>X<EFBFBD><58><EFBFBD>c<><63>G<EFBFBD><47><EFBFBD>pY<70>&<26>Zxƌ<78>wiB<69><42>g<EFBFBD>-<1A>Ƭ<c<><63>ɴ +<2B>X`)<02>Ϙ<>$J<><4A> <c<>Y0<15>J4&<26>=ϬslDc<44>X<EFBFBD><58><EFBFBD><EFBFBD><12><05><><EFBFBD><EFBFBD>F4f<34><66>zE+Ϙ<><CF98>><3E><><16>1c<31>m<EFBFBD><6D><EFBFBD><EFBFBD>s=<3D><><EFBFBD>g<EFBFBD><67>Ɍm<C98C>~-<cƮҠ<C6AE>f<EFBFBD>P<11><>hƌ<68>a\Tb<0F>1a<31><61><EFBFBD><EFBFBD><EFBFBD>f,0<><30><EFBFBD>ߵ<EFBFBD><DFB5><1B> <1D>͘<>~<7E>b<7F>3<><0E>^x<>iƌ<69><C68C>G<<3C>B3f<33><66><EFBFBD><EFBFBD><EFBFBD><EFBFBD>f<EFBFBD>Xv[<5B><><EFBFBD>f̘\<5C> hƌ<68><C68C><EFBFBD>߹Ќ c[EWWo4c<34><63><EFBFBD>i<EFBFBD>ь<19>zeY<65>Ҍ<19>$<24><17>f<D883>X<EFBFBD><03><>4c<34>Α]<07><><EFBFBD>c<><63><EFBFBD><EFBFBD><EFBFBD> <c<><63><19><><EFBFBD>3f<33>QQ<51>1<EFBFBD> cJ<63><4A>>0<>#<23><>yWV<57>1c<31><63> <1E><>4f Xp<58><70>Lc<4C> <1F>ƌ<EFBFBD><C68C>XUؘƌ 3<><33>ycq5<71>@<40>Lc<4C>x<EFBFBD>z<EFBFBD>W<EFBFBD>1c̳<63>=<3D>2<EFBFBD>S<><53><EFBFBD>^ Ә1<18><>e<EFBFBD>R<EFBFBD><19>)d<><64>4f<34>QӼ1<D3BC> ;i<><69>LcƠ<07>m<EFBFBD>F5f<35><66><EFBFBD><EFBFBD>X[<5B><> <15>$<24><>LcF<63> <20>E<EFBFBD>2<EFBFBD>Á<><C381>B<EFBFBD><42>4&(;-<2D><><EFBFBD>4f c<><63>Z<EFBFBD>ƌ<EFBFBD> <0C>Ә1Hew<65>٘ƌ<D998><C68C><EFBFBD>k<EFBFBD>3<>H<EFBFBD><48>xƄA<C684><41><EFBFBD>W<>1c<31><1A>E&<26> 3Fh^$d<0F>1c,<01>`Y<>ƌQ<C68C><51>M<EFBFBD><12> <09>-<08><><EFBFBD> Dcƚ<63>s<EFBFBD>v<EFBFBD>1c<31><63><EFBFBD>W<EFBFBD>1CX''mlDc<44>Xr<58><72><EFBFBD><EFBFBD>5<>L9<<3C><11>s<><73><EFBFBD>w<EFBFBD>1C<31><43>Ѓh<D083>X<EFBFBD><58>%W<>1A
|
||||
N<EFBFBD><EFBFBD>j%3<><33>8<0E>J4f g<>с<EFBFBD>h<EFBFBD>3,b<>B4&,9<>-Dcƺb<><62>Dc<44>¬xۉƌ1F<31>=<3D><10> b <1F>`*]<5D>ƌ<EFBFBD><C68C><EFBFBD>U<EFBFBD><55>h<EFBFBD>Pq3<71><33><EFBFBD>h<EFBFBD>3u|~<16>1A<31>) <1F>ƌ);<3B>b<13><><EFBFBD> a%<25>Օ<EFBFBD>h<EFBFBD><18>><3E><> <09>)<29><><EFBFBD><EFBFBD>g<EFBFBD>s<><73><EFBFBD>%<25><>3<><33><EFBFBD>3<EFBFBD>Ϙ<>Ӧ<EFBFBD><D3A6><EFBFBD>g<EFBFBD>$<24>qQ<71>=h<> E7η<37>fL<66><15>}<7D>iƌ<69><C68C><EFBFBD>|<7C>3<><33><EFBFBD><EFBFBD>͘0\<5C>8= hƌU <0C>B<EFBFBD>4c<34>ҕ+<2B>ь <09>ph!<21>V<EFBFBD>1c<31><63><EFBFBD>+͘1<CD98> [hƈ<68>dW<11><><EFBFBD>f<EFBFBD><66>y<EFBFBD>'l4c<34><63>#<23><>ь<19>}<7D> g<D883><18>e'<27>lDc <63>]<5D>h%3&Sj-Dc<44>><3E><><19>)<29>s<EFBFBD><11>;<3B>eư<65>iLX}X<>Ә<>|<7C><>nLc<4C>
|
||||
<EFBFBD><EFBFBD><EFBFBD>j<EFBFBD>2<EFBFBD>{t<>ܘƄ<DC98><C684>z7C<37><43>4f<34><66>Khe<13> <1D><15>0<EFBFBD>; <1F><>W<EFBFBD>1a,J:M<><4D>2<EFBFBD><19><><EFBFBD>b{0<> <09>9<EFBFBD>ݵ<11><19><<3C><>ۈƈ<DB88><C688>rr+<2B><>h̘<68><CC98>6<EFBFBD>1Á?a ˘1<CB98>D<EFBFBD>0deF<07> <09>V<EFBFBD>1CL\
|
||||
R<EFBFBD><EFBFBD>fLX+WGȍfL;8<><38>ь#Ek<0F><>2&<26>I<EFBFBD><11> ;<3B><07>`3<>XJ`<0F>1aYF<59>oa3v\<5C><><EFBFBD>4&,=<3D>36<33>1ȃt<CC91><74><EFBFBD>4F<34><46><EFBFBD><EFBFBD>K6<>1cL<63>-<17>jL<6A>`s<><73><EFBFBD>B5&<26><>6<EFBFBD>16<31><36><EFBFBD>\<5C>Ƅ5<C684>N<EFBFBD><4E> \c<>Ȕ,+٘<><D998>]<5D><><EFBFBD><EFBFBD><EFBFBD> <09>{h<>X<EFBFBD><58>ƌ1<C68C><31><EFBFBD><EFBFBD><EFBFBD> dc<64> <20>#<23>ic<13>"<22><>mgƬ<>3h<33><16>1b<31> }<7D>7<EFBFBD>1a<31><04>ҹ<EFBFBD><D2B9>;<3B><0F><>6&<26><><EFBFBD><EFBFBD><EFBFBD>-<2D>B7&<26><><EFBFBD>7<EFBFBD>1a<31><61>u<EFBFBD>U<EFBFBD>tc<74> <20>[<10><>tc°<63><C2B0><EFBFBD><EFBFBD><EFBFBD>nL<18>SGP<47>-tc<74> <20>̑<EFBFBD><CC91><EFBFBD> a'<27><>EY<45>ƒV<C692>z<EFBFBD>Y]V<>1a<31>l
|
||||
<EFBFBD>"<13><><EFBFBD><EFBFBD><EFBFBD><EFBFBD> c<><63>i<EFBFBD>W<EFBFBD>1Al<41>?.<1A>ט<><D798>M<EFBFBD><4D><EFBFBD>+ט<><D798><EFBFBD><EFBFBD><EFBFBD>i<EFBFBD><13><>S<EFBFBD>W<EFBFBD>1 <0B><><EFBFBD>4F<34>K<EFBFBD>cE<16>1A4<41><34>MnLc<4C><63>~<7E><><EFBFBD>[<5B>Ƅ<EFBFBD><C684><EFBFBD> <0C><><EFBFBD>iL<18>pP<14>Tc <03><>zY<7A>Ƅ<EFBFBD>w<><EFBFBD>ԫaV<61>a۩<7F><04>y郼P<E983BC> b<><62> <09>V<EFBFBD>17<37><7F><EFBFBD><EFBFBD>4&<26> $g<><67>4&,<2C>U<EFBFBD>O<EFBFBD><4F>4<EFBFBD>V<13>t<14>0<EFBFBD><30>EK<><4B>
|
||||
we<13>e <0C><> Lc<4C>lљ b<><62><EFBFBD>iLY<>l<EFBFBD>lLc<4C>pa "oLc<4C><63><EFBFBD>^<5E><>:p<><70><EFBFBD>sY<73>Ƅ<EFBFBD><C684><EFBFBD><EFBFBD><EFBFBD><16>1w<31><77>Q:{P<> <09>-<2D><>g<EFBFBD><13>L0<4C><30>TcJ<63><4A>b`<60>@5<><35><EFBFBD><07><>B5<42><35>#<23> <1E>7<EFBFBD>R<EFBFBD> <09><0F>Ũw<C5A8>1w39<15> <0C>A5<41>.(3<>_<EFBFBD>6<EFBFBD>1AW2<57><32><EFBFBD>5<EFBFBD><35>*
|
||||
d~<7E>s?<3F><16>F5<46>6.<2E>Rt<52><1A><>08<30>~;+[<5B><14>e<EFBFBD><65>+<>Q <1D>L<EFBFBD>4W<34>1wũ<77><C5A9><EFBFBD>Ә<><D398><EFBFBD>7<EFBFBD>ؘƄU<06><>r<EFBFBD>2<EFBFBD>%<25>un<75><6E>N5<4E><35>W<EFBFBD>괳Q<EAB4B3>EC<45><11>>[1<><31>bc<62><13>ؒ-<2D><> \cn>D<>t<EFBFBD><74>5&<26>!<21><><EFBFBD>u<EFBFBD><1A><>F<EFBFBD>g<EFBFBD>ט<> щ<>&<26>٘;$=<odcn<63><6E><EFBFBD>%<25><><EFBFBD>lL<6C><01>(TЕlLYp<59>8<EFBFBD><38><EFBFBD>F6&<26><><0F>d%s<07>G<EFBFBD><47><EFBFBD>6<EFBFBD><36>Q<EFBFBD><51><EFBFBD>`s7<73><37><EFBFBD>8<EFBFBD>[<5B><><EFBFBD><EFBFBD> <0B><><EFBFBD>=<3D><><EFBFBD> <09>2 ,la<13>.<2E>A><3E><><EFBFBD>k<>h<EFBFBD>{z<><7A>Ƅ9q=<07>`<13>|<7C><11>i <0B><18>Z<15>0<EFBFBD><30><EFBFBD>6&<26>%D<><44><EFBFBD><EFBFBD><EFBFBD>*W'ʷ<>lLu4<75>ٕk<D995> ){<7B><>4&<26><><EFBFBD> laS<>#<23><>ܝdcF5.<2E><><EFBFBD>iL<18><><EFBFBD><EFBFBD>^<5E>ƈ<EFBFBD>{9<><39>ka<13><><EFBFBD><EFBFBD>ݬLc<4C><0E>O<EFBFBD>t<>c<>&<26><11><a 0<D598> <1E><><EFBFBD>j<EFBFBD>-<2D>I<EFBFBD><15><16>1a'f.Z<>ƈa^5<>57<35>1aj<61><6A><EFBFBD><EFBFBD>+<2B><18><><EFBFBD><EFBFBD><EFBFBD>4~%F<><46>i<02><>lLɩ<4C><C9A9> <1F>ƈ<EFBFBD><C688>W{<7B><> s<16>I<EFBFBD><16>1bg
|
||||
[<5B>ƈefh\<5C>\<5C>Ƅ<EFBFBD><C684>)hV<>1b)<29><><EFBFBD><EFBFBD>l<EFBFBD><18>벐<> #E}<7D>6<EFBFBD>dcL<63>ś=$K<>v<EFBFBD>1%<25>r<EFBFBD>^<5E>Y <0B>XR<58><52>!6<>1a<31> (7<>_<EFBFBD>ƈ<EFBFBD><C688>C<EFBFBD>Gz##F<>$<24><>acgC<67>>
|
||||
o<1B><18>S<EFBFBD><53>O<EFBFBD><4F>B7&,E<><45><EFBFBD>F7F(S<>:H<>э)˙
|
||||
<EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>ƈQ4<EFBFBD><EFBFBD><EFBFBD>m<EFBFBD>c<><63>d<EFBFBD>t<EFBFBD>6<EFBFBD><36>ƈqߨO<DFA8><4F>N7FL<46><4C><EFBFBD><EFBFBD>@7<>T7<54>v))<29><07>1<>=<1A>07<30>B7<42>,<2C>6$O<>-tc<74><63><EFBFBD><EFBFBD>O<EFBFBD>;<3B>!n<><6E><EFBFBD><EFBFBD>F7F<37><46><EFBFBD><EFBFBD>_<EFBFBD><5F>Ӎ<EFBFBD>7<EFBFBD><37><EFBFBD>v<EFBFBD>c<><63><EFBFBD><EFBFBD><EFBFBD>. +<2B>!<21>K<EFBFBD>#<>S<><53>Q <1C>y[<5B>ƈT<C688>m<EFBFBD><6D><EFBFBD>lc<6C><63>/<04><><EFBFBD>_<EFBFBD>ƈ<EFBFBD>sJ;<3B>} <0B>!֤<>#Hl%S<16><>*6<>1&<26>c:<3A><>m<EFBFBD>KzL<7A><4C><EFBFBD><EFBFBD><EFBFBD>\<5C>ƈ<EFBFBD>Zd<5A><64><EFBFBD>.<2E>1<EFBFBD><31><EFBFBD> <1D><><EFBFBD><16>1bL)f{da<0F>1"<22> <1D><>Ry<><79>%<25> <0B><><EFBFBD> <0C>ҍ<EFBFBD><D28D> L<><4C>
|
||||
<EFBFBD>N7<EFBFBD><EFBFBD>DŽt?<3F>B7<42>rn<>q!<21>lc<6C><63> <1C><>OygS}ƮMc<4D><63><EFBFBD>eUN]<5D>J;<3B>XVq8<71>B<EFBFBD><42>F7F c<><63>˵э<CBB5><D18D><EFBFBD> ,<2C><><EFBFBD>Ӎ<EFBFBD>i<02>R<EFBFBD><18>ҍ<EFBFBD>l01O<31><4F>(<28><><EFBFBD>Ʋ<EFBFBD>c<EFBFBD><63><EFBFBD><EFBFBD>ݘh<05>oL<6F><4C>ҍe<D28D>=s<><73><EFBFBD>N7<4E><37><1A>i<EFBFBD><69>ʕn<CA95><6E><16>v<06><>J7<4A><37>YP<59>#A+<2B>X<EFBFBD><58>:Y<>f갅p<EAB085><70>:ֲ<>s<EFBFBD><73>c$S<><53><EFBFBD>(S<><53><EFBFBD>c<EFBFBD>>
|
||||
r<EFBFBD><EFBFBD><EFBFBD>e# ;{<7B>X<EFBFBD>o<EFBFBD>c<EFBFBD>n`<60><><EFBFBD><EFBFBD>V<EFBFBD>1`<60><>.<2E><><EFBFBD> <09><><EFBFBD>><3E>'w' #SDR=@ <09><>p<EFBFBD><70><EFBFBD><EFBFBD><EFBFBD><EFBFBD>f<EFBFBD><1B>ӓ[<><EE9BB3><EFBFBD><EFBFBD>&t-v<0E><>o<EFBFBD>M<>d6<64><36><EFBFBD>7F<37><46> k'<15>|c<>,<2C><>x<04><><EFBFBD>7F<37>-<2D><><EFBFBD>&+ݘ<><DD98><EFBFBD>Q<EFBFBD>/<2F><>n<EFBFBD>IY <0C><>
|
||||
<EFBFBD>э<EFBFBD><EFBFBD>z<><7A><EFBFBD><EFBFBD>lc<6C><63><EFBFBD><EFBFBD><EFBFBD>Mf"[<5B>ƒb<C692><62>Ty<54><79>6<EFBFBD>~+<2B><><EFBFBD><EFBFBD>b"{gKڤ<4B><DAA4><15><><EFBFBD><EFBFBD><EFBFBD> Vqz<71><7A>6F9S<39>Ms<4D><73><EFBFBD><EFBFBD>/ <0C>,<2C>*<2A><>6<EFBFBD>t<EFBFBD><74>L<EFBFBD>6gcc<>[<5B><><EFBFBD>`g<1B><01>)Է<>m,)s5<73><35><EFBFBD>ƒ<EFBFBD>ұٿ<D2B1><D9BF>V<EFBFBD><56><EFBFBD><EFBFBD>\<5C><> <0C><>m,)<> :f<>Z<EFBFBD><5A><18><>
|
||||
<EFBFBD>p<EFBFBD><EFBFBD>6<EFBFBD>X<EFBFBD><EFBFBD>^C~<7E><>m,<2C> /<2F><><EFBFBD><EFBFBD>m<EFBFBD>vu=Ώlc<6C> د<>E;ۘ<><DB98><EFBFBD><EFBFBD><EFBFBD>D<EFBFBD><44>6Fݑ<46><DD91><EFBFBD>pcc
|
||||
&~<7E>0_<30><5F>6<EFBFBD>:Kv<4B>L<EFBFBD>_7<5F>15&<26>;<3B>#<23><18>_z\<5C><><EFBFBD><EFBFBD>)pu@t<><74><EFBFBD><EFBFBD><EFBFBD>m<EFBFBD>٧\\co<><6F>y<EFBFBD>;<3B><18><><EFBFBD>b<EFBFBD><62>J6<4A>1<EFBFBD><31>b<EFBFBD><62><EFBFBD>mLc<4C>Jv<4A><76><EFBFBD>4v(b<><62>@<40>J5ƪ[<5B><1A><>B<EFBFBD><42>4F<34>P<EFBFBD>!-~<7E><>i<EFBFBD>iÓ%V^<5E><>i<EFBFBD>N<><4E><<3C>>و<>h<EFBFBD><68><EFBFBD><EFBFBD>=9<1B>ۄ<>O<EFBFBD><4F><EFBFBD><EFBFBD>2<EFBFBD><32>f: <1E><>u%c`<60>:<3A>/<2F>A2<41><32><EFBFBD><EFBFBD>lR<6C><<3C> cS<63>H܄ (Ʀh$ToЃal*<2A><><EFBFBD>[<5B><>Z<08><><EFBFBD><EFBFBD>0<EFBFBD><30><EFBFBD>r<EFBFBD><17><>v<EFBFBD><76>Mo<4D><6F>^L<>ߙo<DF99>><3E><>1_<><5F><EFBFBD><EFBFBD>Vr1<72><31>S' &<26><>\<5C>/<2F>~#<23><>n<><6E><EFBFBD><EFBFBD>s<EFBFBD><73><EFBFBD>Z<EFBFBD><5A><EFBFBD>^<5E><> v+n<>b<EFBFBD>D<EFBFBD> <1E>L<EFBFBD>r<EFBFBD>
|
||||
1M<EFBFBD> <04>+<2B>{<14><><EFBFBD> g <0B><>Z+ <1C>sAj1ud<75> q<><71><EFBFBD>[<5B>}Y$<24><>n1<6E><31>.g<><67><EFBFBD>-<2D>Ό&<26>6?<3F><>-<2D>I<EFBFBD>Av<41><76><EFBFBD><EFBFBD>-<2D><><EFBFBD><EFBFBD>if <1F><>8<EFBFBD><38><EFBFBD>D<EFBFBD><1B> <0B>ϼ<A <0B>g<>i<EFBFBD>=<08> <1E>blP<6C>?<3F><><EFBFBD><EFBFBD>-F? 7<><37>
|
||||
7r1r6@<40><><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>qyσ}<17> c<11><>5j%<17><><EFBFBD>6<EFBFBD>{<7B>o<EFBFBD>b<EFBFBD>w<EFBFBD><77>~<7E><><EFBFBD><EFBFBD>[ <1D><><13>;<3B><>|<7C>
|
||||
6 Z<><5A>&<26>l=l<><6C><EFBFBD>^Lْ<4C>Z@<40><>bnCz<43><7A>|<7C>c<>G<EFBFBD>k<EFBFBD><6B>Vz1Q6<51>1<EFBFBD>l<>+<2B><04><>q<EFBFBD><71>^<5E><><03>'<0E><><EFBFBD><EFBFBD>b4E<34>Xf"<22><><EFBFBD>h<EFBFBD><68>Nu<4E><75><EFBFBD>/Fk<46>il<69>V~<7E><><EFBFBD><04><><EFBFBD>=<3D><>5z<35>`<60>o<><6F>/&<13><>(oO<6F>b4NIdj<64><6A>_<EFBFBD>)<29><><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>o<EFBFBD><6F>`<60><>;<3B>p#<23><>`<60><>k<16>
|
||||
<EFBFBD>=<11>u<EFBFBD><75><EFBFBD>;<3B>Xg?>5<><13><12><><EFBFBD>@g<><67>"<22>N0<4E>{<1B><>{<7B><>c<>M7Z
|
||||
ޟ<EFBFBD>`<60> <1D>N<EFBFBD>^<5E><>`l<><6C>Ճ7m!<1B>i<EFBFBD>#^<5E>F06<30>-<2D><>>A+<2B><18>W<EFBFBD><57>x<EFBFBD>c<><63><EFBFBD>W<EFBFBD><57><EFBFBD>`<60>:<3A>ن<EFBFBD>4;<3B>ؼ<EFBFBD><D8BC><EFBFBD> <09>6~<7E><><EFBFBD>}ت)<29>w<EFBFBD>1<EFBFBD>xOܼ<4F><16>1х@7<18>B+<2B>o?<3F>R<EFBFBD>#i#Se<16>?<3F><>l!c<>h<><68><06>F0&<26>6H<=j!<18>=I7[<5B>i,c<><63>b<EFBFBD>dSVlcl<63>B<EFBFBD><42>`[[ <09><>g<EFBFBD><67><EFBFBD>+<2B><07>s<>H<EFBFBD> <1C>N+<2B>o<> <0C><> <0C><><EFBFBD>`L}X3<58>W-<2D><07>ou<6F><75><EFBFBD>M<EFBFBD><12><><EFBFBD>,<2C>{~ c<><18>;<3B>P<1B><>XUآ<55>˻<12>1AN
|
||||
D7<EFBFBD><EFBFBD>B06O<EFBFBD> c; A <0B><18><><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>b<EFBFBD><62><EFBFBD>{3<>m<><17>j<EFBFBD><6A><EFBFBD>r^$b~1v<31>a<EFBFBD><61>
|
||||
<EFBFBD><EFBFBD>_l<EFBFBD><EFBFBD><10><>Z<EFBFBD><5A>_<EFBFBD>Iz<49>"<22><EFBFBD><F18B91B0><EFBFBD>=<3D><><16>1<EFBFBD><31>e <1C> ] <0B><18>
|
||||
<EFBFBD>e <1D><>\<5C><>X{<7B>x<EFBFBD>a.<2E><>_<EFBFBD>mty<><79>l!<13> =<3D>f \ <09><18><><EFBFBD>W<>J0<4A>b><3E><><EFBFBD>z<10><>/Oe<4F>}<<3C>B0FE+<2B>8<EFBFBD>$A+<2B> <1F><>?<3F>ö<10><>8<EFBFBD><38>*<2A> k!<18>2<EFBFBD>fAZ`<60>Mx<19>;|<7C><>cT<63><54>1<EFBFBD>
|
||||
7<>1<EFBFBD>X<EFBFBD>b<EFBFBD><13><12>1<EFBFBD><31><19>nO<6E><11><>D<EFBFBD>E<EFBFBD>3<EFBFBD><07><18><>=h{v%c <1F><><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>`<60>-<2D><>-<2D><>`LHe><3E> j!<18><>I<EFBFBD>1<EFBFBD><10>q<12><>_<EFBFBD>Qm!<18><>aT7^<5E>ƸL<07>X<EFBFBD><16>1<EFBFBD>-<2D><><EFBFBD><EFBFBD>`l<>?<3F>~j<>+<2B><><EFBFBD>ly<><79>V<EFBFBD>1<EFBFBD><31>Uoo;<3B><><14>+<2B><>y<><79><EFBFBD>mT<6D><0E>m<EFBFBD><6D><EFBFBD>Ѥa=_(<28><>b3<><33>l<EFBFBD>E"<22><>S<><53>
|
||||
B<EFBFBD><EFBFBD>V~1mG<6D><47><EFBFBD>#<23><18>=L<><4C>_L<5F>q<08><01><><EFBFBD><EFBFBD>+t <09>w<EFBFBD><17><><EFBFBD><EFBFBD><EFBFBD><EFBFBD>m<EFBFBD><6D>b<EFBFBD><62>N<EFBFBD><4E><05>N/F˛<46>o<EFBFBD>ﻍ^<5E><>}c:<3A>e<EFBFBD>F/FY.<2E><>ޞ<EFBFBD>ņ<EFBFBD><C586><EFBFBD>.q<>c<>(U/<2F><><EFBFBD>/<2F><>
|
||||
<EFBFBD>?<3F>W<12><><EFBFBD><EFBFBD>b<EFBFBD><62>|0]<5D><>b*<2A> <20><>ˣVz1Ve<56>K<EFBFBD>y<EFBFBD><79><EFBFBD>X]<5D>b<07>iVz1
|
||||
<EFBFBD>c<EFBFBD>߸Ћq"ʔ<>%<25><><EFBFBD>q 5<><35><EFBFBD>vr1b<><62><08>\<5C>m<>&<26><><EFBFBD><EFBFBD>b,9߹ Vr1Ξ-N<><4E><EFBFBD>F.F<> <20><>i<EFBFBD>h%㣩_B\<5C>+<2B><18><><EFBFBD>:Rh<52>+<2B><>*fX-?MP<4D><50><EFBFBD><EFBFBD><EFBFBD>x!/<2F><><EFBFBD>.<2E><>Cϼs/7v1~'V<><56><EFBFBD><EFBFBD>[<5B><>8<EFBFBD><38><EFBFBD><EFBFBD><EFBFBD>.<2E>g8)z.la㒕 (<08><>.<2E>7<EFBFBD>T <1F>ň<EFBFBD><C588><EFBFBD><EFBFBD>j6<6A><36>]L<>9ɴ.}vc<17>+<2B>%x<>hesM<73>m<EFBFBD><6D><EFBFBD>\<5C><><EFBFBD>֒<EFBFBD>=<3D><07><18><>V*<2A><><EFBFBD>r1.<2E>d<EFBFBD><64>;<3B>l<EFBFBD>bZ<62><5A>
|
||||
<EFBFBD>J.f<>p<EFBFBD>䷝\L<><4C><EFBFBD>&%[<5B>Ŵ<EFBFBD>'<27><>LX<4C><58><EFBFBD><EFBFBD>2 <1D><> <17><> <1F><><EFBFBD> k[<5B>ň<EFBFBD>(<28>B<EFBFBD>\Z<17>ײ<>gw<0E>A.<2E>=m<>X<EFBFBD>;<3B><>^ <1D>̎ +[<5B>ń<EFBFBD><C584><EFBFBD><EFBFBD>kv1}<7D>e<01><><EFBFBD>b<EFBFBD>Ik<49>A{<7B><><EFBFBD>t<EFBFBD><74><EFBFBD>:<3A><> <0B>1<><14>
|
||||
<EFBFBD>A/<2F><>EG<45><47>b7z1I7v<37>3<EFBFBD><33><EFBFBD>.<2E><><EFBFBD>h<EFBFBD><68><EFBFBD><EFBFBD><EFBFBD><EFBFBD>%<25><>5<13>-<2D>b<EFBFBD><62>)ֽ( <0B><>ՌVlo;<3B><><EFBFBD>EZ<45><5A>N.<2E><>ÿT#<23><><EFBFBD>]Lg<4C><07><><EFBFBD><EFBFBD>.F<>u!6v1m"<22><>1<EFBFBD><31><EFBFBD>.<2E><><EFBFBD>^<5E><><1B><><EFBFBD><EFBFBD>tK`v1<76>=h7<68>R+h!<17><16><><EFBFBD>y<EFBFBD><79>ň<EFBFBD>-x<>vr1<72>[<5B><>a<EFBFBD>j#<17> <1E><>q<EFBFBD>Wr1<72>cp<63><70><EFBFBD>[<5B><>$<24>H<13>"<22><>\Le<4C><65><EFBFBD>i <1F><>$<24><>G-<2D>F.<2E><>I<EFBFBD>:<3A> <1F>Ŵb<C5B4><62>`<60>ыy<D18B><79>$<24><><EFBFBD>^,<2C><>G<EFBFBD><47>MK<4D>Ћ<EFBFBD>Ͳ<EFBFBD>F:n<>b l{RM<52><4D>ҋ<11><>|e<>m<EFBFBD>b<EFBFBD>ed<65><64>N<EFBFBD><4E><EFBFBD>Ŵ;+{<7B><><EFBFBD>k<EFBFBD>Ӯf<19><15><>_L<18><04><>l<EFBFBD>b<EFBFBD><62>C<EFBFBD> <08>t<EFBFBD>]<5D>&<26>Z<08>t| \,b<0F>1<EFBFBD>Ȏ<>_<EFBFBD>c<0F><><EFBFBD><EFBFBD><EFBFBD><EFBFBD>p(g#<18>r<0E><><EFBFBD>4W<34>1<EFBFBD>"<22>43<34>J0<4A><30><EFBFBD>4$w<><77><18>$<24>X<EFBFBD>ZL <1D>2<EFBFBD>Iϙ<49>U<EFBFBD>v<EFBFBD>1 "<22><19>#V<>1<EFBFBD><31><EFBFBD>[<5B>mg#<23>;<3B><>,<2C><16>1 0<><30><EFBFBD><EFBFBD>V<EFBFBD>1_͍U3<55>ʅeLg<4C>N}c<><63><EFBFBD>Y R5<52>E@<40>`<19><>%<25><><EFBFBD><EFBFBD>2&<26>?<3F>]U<><55>2<EFBFBD><32>7<EFBFBD><37> <0C>b#<23>T2<13><>2&= _f<><66>2<EFBFBD>S =<3D><>V<>1<EFBFBD><31> <0C>x<EFBFBD>W<EFBFBD>1<EFBFBD>:<3A>y<EFBFBD><79>p%<19><03><><16>1<EFBFBD>3 7<><37><EFBFBD>cR<63><52> ō<>6<EFBFBD>1]A<><41><EFBFBD>=<3D>F1<46>={<7B><07><><EFBFBD>1<EFBFBD><31> $,<2C>V<EFBFBD>1<EFBFBD><31><EFBFBD><EFBFBD><EFBFBD><EFBFBD> c:搦-<2D><>V<EFBFBD>1_BL(4<><34><EFBFBD>1<EFBFBD><31><EFBFBD>"<15><><EFBFBD>8ƈ<15><><EFBFBD>w<EFBFBD>q<EFBFBD> c<>D3<44>o c w:<3A><><EFBFBD><EFBFBD><EFBFBD><EFBFBD>X <09>n<EFBFBD><18>4<EFBFBD><34><EFBFBD><EFBFBD><EFBFBD>dL<64><4C><EFBFBD><EFBFBD>M<EFBFBD>7<EFBFBD>1+\VՄ-$c١ywHy<48>IƤ
|
||||
<EFBFBD><EFBFBD>$<24><><EFBFBD>dL:#V(<28>`<60>ZH<5A>,<2C>y<EFBFBD> <1C>$cֲ<19>4[<5B>J2&<26><03>2<EFBFBD><32><EFBFBD>J2<4A>Bx<42>FXq<58>HƤ<48><C6A4><EFBFBD><EFBFBD>v(,,cR<63><52>*Ԭ_ ˘<><CB98>5 c ˘<>O
|
||||
v<EFBFBD><EFBFBD><EFBFBD>e,<2C><>]U
|
||||
<EFBFBD><EFBFBD><EFBFBD>e<EFBFBD>b<EFBFBD>-<2D><>-la<19>Q<EFBFBD><51><EFBFBD><EFBFBD>˘$@a<16><><EFBFBD><EFBFBD>eL"l<>ɢ<19><16>1ɰo<>m<EFBFBD>6<EFBFBD>1a<31><61><EFBFBD><EFBFBD>fL<66>ʉ-f<>Z<EFBFBD>Ƅ1<C684><31>F<><46>hL<68>A<EFBFBD> <1D><><EFBFBD>hL<68>W<EFBFBD>1u<31><75>l!<1A><><EFBFBD><16><10><><EFBFBD>ȹk<C8B9><6B><EFBFBD>fL<66>}Ў<>]<5D>Ҍi5S<35><53><EFBFBD><EFBFBD>4cz <0B>)g<>i<EFBFBD><19><><EFBFBD><EFBFBD> #v<><19><>
|
||||
<EFBFBD>)<29>ޕg<DE95>U<>j<><6A>3F<33><46><EFBFBD>3<EFBFBD>'W<>1 <09>:<3A>g<EFBFBD><67><cڷ<63><DAB7>-<2D><><EFBFBD>i<EFBFBD><69>_j<5F>oϘ<>-<2D><><EFBFBD><EFBFBD><EFBFBD>3&ݖ&y<>]<5D><><EFBFBD><EFBFBD><EFBFBD><04>h\-<2D><11>Yϫ<>t%<1A>1<EFBFBD>8<EFBFBD>6<EFBFBD>1<EFBFBD>[W<><57><EFBFBD>N4f<19>Ԧ<EFBFBD>{<10>I<EFBFBD>8G<38><47>\<5C><>,<2C> &C<0E>Dc<44><63>L<b<> ј <02><><EFBFBD>.<2E>A5&<26><><EFBFBD><EFBFBD>"X<><58>,kS;<3B><>ۨ<EFBFBD>,<2C><><EFBFBD>+Ә ֔;,<2C>1<EFBFBD> ;<3B><><EFBFBD><EFBFBD>-Lc<4C> <1E> 6<>1aE<61>z̈<7A>0<EFBFBD><30>*a <1F>`D[<5B><><EFBFBD>9<><39><EFBFBD><EFBFBD>4&l\ R7<52>1<EFBFBD>J<EFBFBD>Fo;јV<D198><56>ENiۈƬ,<2C>O<EFBFBD>{<10><><EFBFBD><EFBFBD>{ 7<>1<EFBFBD><31><EFBFBD><EFBFBD>(<28><>Z<EFBFBD>ƴ[(4<>4ga<1A>
|
||||
tY<EFBFBD>јt<04><><EFBFBD>3Ϙ̿<CF98><CCBF>;<3B>j<EFBFBD><10>i<EFBFBD><69><EFBFBD>,zۘ<7A>∐<EFBFBD><E28890>߸0<DFB8><30>d<EFBFBD><64>3e<33>B5<42>{<7B>4I.<2E>ب<EFBFBD><D8A8>lS<6C><53>G<EFBFBD>1_j<5F>;<3B>r<EFBFBD>TcĘ<63><C498>NR<4E>=<3D><>(8<><38><EFBFBD>#,Tc<54><63><EFBFBD>[<5B>oTcTԨ<54>{<7B>Q<EFBFBD>Y1?|<7C>mDcr<63>@<40>
|
||||
'<27>lDcR<63><52>xp<78><70>Dc<44>A<EFBFBD>2*<2A>[<5B>Dcv3<76>C<>J4v<34><76><EFBFBD>$<><7F>4&c<>v|
|
||||
<EFBFBD><EFBFBD>4<EFBFBD>+<2B><><EFBFBD>\o<>1<EFBFBD><31>*q<>
|
||||
<EFBFBD><EFBFBD>4<EFBFBD><EFBFBD>wZr<05>`<1A><>Apa Ә.<2E><>d
|
||||
<EFBFBD><EFBFBD>-Lc<4C><63><EFBFBD>g<7F><67>4拲eu<65>{ۙƄqK8 <1E>2<EFBFBD> <09>A
|
||||
}kc<1A><>̪<EFBFBD> <20>zg3ğ<33>4:<17>1<EFBFBD><08><><EFBFBD><EFBFBD>4<16>j<16><><EFBFBD> Lc<4C>x)<29><>ޙ<EFBFBD><DE99>3<EFBFBD>
|
||||
<EFBFBD><EFBFBD>iLXV ο<>0<EFBFBD><30>®7<C2AE>Ә<><0F><>5˅i<CB85><69>5<EFBFBD><35>\ee<13><>
|
||||
<1F>ƴW9<15><> Ә<><D398>%<25><>v<EFBFBD>1<EFBFBD><31><EFBFBD><EFBFBD>^><3E>+ӘT <0B><><EFBFBD>%Ә<>Ba<42><61><EFBFBD>Ә<><10><><EFBFBD>B<EFBFBD>1<EFBFBD><31><EFBFBD>]<5D>lo;ӘT-<2D><>ke<1A><><17><><EFBFBD>v<EFBFBD>1<EFBFBD><31><EFBFBD>T<EFBFBD><54> Lc<4C> WG<57><47><EFBFBD>iL<69>LV t3<74>-Lc<4C><63>C<EFBFBD><43><EFBFBD>re<1A>$<24><>5L<35><4C><12>ɦ|o#<23><11>Ip<49><<3C><><EFBFBD><EFBFBD><13>9l ?E<0E>A4&a 3<>C<> Dcr+<2B>j9ؽ<16>1Y% <0B><><EFBFBD>3<EFBFBD>K<EFBFBD><4B>[<5B><>k<EFBFBD><19><>S"U<>m<EFBFBD><19>J<EFBFBD><4A>ɦ<EFBFBD>Zy<5A>t=eV<65><56><EFBFBD>t<EFBFBD><19><><EFBFBD><EFBFBD>&l<c<><63><EFBFBD><EFBFBD><0E><17>gL<67>X<EFBFBD>8<EFBFBD>7<EFBFBD>1<EFBFBD>s<EFBFBD><15>4=<3D>B3<42>x<EFBFBD>U;<3B><><EFBFBD><EFBFBD><EFBFBD>.<2E>b<EFBFBD><62><EFBFBD><EFBFBD><EFBFBD>f<EFBFBD><66><EFBFBD><EFBFBD>w%<19>
|
||||
0<1A>><3E><><EFBFBD><EFBFBD><EFBFBD>,A?<3F><><EFBFBD>I<EFBFBD>p?z<>i-$c<>l<EFBFBD>p<EFBFBD><70><EFBFBD>1&<26><> <13>n c<><63><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>1f<31>p<EFBFBD>(<28><>HƬ4<C6AC>Օ<EFBFBD><D595>eL<18><>Eɘ<>D<EFBFBD>&S<><53>$c<><63><EFBFBD>è<EFBFBD>HƄEw<45><77><EFBFBD>c<EFBFBD>*a<>Q*<2A>q<EFBFBD> <09>D/h<><13><>n<EFBFBD>[<5B>r<EFBFBD> <1F><><EFBFBD>c<EFBFBD>٢<> ǘ<>vho c<><63><EFBFBD><EFBFBD>[(<28><ɤ<>A <0B>X<EFBFBD><58><EFBFBD>~<7E>(Ƭ^<5E><>߭cV<63>ULdB<64>w<EFBFBD>1+<2B><><EFBFBD><EFBFBD>[de<19><>\<5C>#\ɘm 6Ʊ<36>ZI<5A> <1C><><EFBFBD><11><><EFBFBD>y<0F><><EFBFBD><EFBFBD><EFBFBD>e<EFBFBD><65>5<EFBFBD><1A><>6<EFBFBD>1<EFBFBD>*q<><71>v<EFBFBD>1k<31><6B><EFBFBD><EFBFBD>mg<19><11>}<06>`<19>V<EFBFBD>z<EFBFBD><7A>6<EFBFBD>1 ;<3B>|<7C><>2f<32><66>ȏ><3E><>ٗVr<56>jl,c<>ƃ<EFBFBD>ac<19><>x<EFBFBD><78>V<EFBFBD>1 <13>~`s@<07> <1D>X+͘<><CD98>;>~<7E><19><>s{<7B><> 4c <09><><EFBFBD>i<EFBFBD>$O<>?U?ЌI <0C>z<EFBFBD><7A><EFBFBD>hƄY<C684>7<EFBFBD>N3&<26>N <1F>05<30>J3<4A>t<EFBFBD><19><><EFBFBD>eL<65>n<>i<EFBFBD><69><EFBFBD>2<EFBFBD>ۇ<EFBFBD> <0E><>,c<><63>|hv+˘<>3<EFBFBD><33> <1F><><EFBFBD>,c6d<36><64>K<EFBFBD>ӍeL7d:<3A><>˘<><CB98><EFBFBD><EFBFBD><EFBFBD>eLwn?<3F>x<EFBFBD><78>2<EFBFBD><32><EFBFBD>`<60>Id6<64>1<EFBFBD><31>Y=o<>ߵ<EFBFBD><DFB5><EFBFBD>8:<3A><><EFBFBD><EFBFBD><EFBFBD><EFBFBD>2&<26>$n<><6E><EFBFBD><EFBFBD>I<EFBFBD>hb65<36>ЌI<15><>z<EFBFBD><18><><02>}TC Ø<><C398><EFBFBD>ߞ(<28>d<EFBFBD>5W<35>ߞ(Ƥ-<2D><>Ҳ<EFBFBD>-cҲ<63><D2B2><EFBFBD><EFBFBD>(Ƅ]yrOc<><63> <1C>-d<><18>m<EFBFBD>#<23><><EFBFBD>bL<62><4C>ek<65>mc<><14>:М<>'<27>1i<31><69>F j+Ř<><C598>N2<4E>cRZy#<23><18><><EFBFBD>9<EFBFBD><39>|<7C>jֲw<D6B2>1<EFBFBD>,H}<19>e ˘t<CB98><74>h<EFBFBD>1BmN<6D>ʘ䟢K<><4B><EFBFBD><EFBFBD>q<EFBFBD>O<z7<15><><EFBFBD><7F>ľ<EFBFBD>Io<49><EFBFBD><7F><EFBFBD><EFBFBD><EFBFBD><EFBFBD> [m2&<26><07><><EFBFBD><EFBFBD><EFBFBD><EFBFBD>L<EFBFBD><4C><EFBFBD>@<40>R<EFBFBD><52>$r<><72>I<>Eg<>/<>D8<44>/<2F> :<3A><><EFBFBD><EFBFBD><EFBFBD><EFBFBD><7F><EFBFBD><EFBFBD><EFBFBD>w<EFBFBD><77>~<7E><><EFBFBD><7F>?<3F>U<EFBFBD><55><EFBFBD><EFBFBD>b<EFBFBD>3<><33>ǟ <1C><>'U<>M<EFBFBD><4D><EFBFBD>/ <1E><>!e<><65><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>~<7E><><EFBFBD><EFBFBD><EFBFBD>|X<><58><EFBFBD>~<7E><>C<EFBFBD>]><3E><><EFBFBD><EFBFBD>E?Ɉ<><C988>>.<2E><><EFBFBD>:,<2C><> <16><>A|<7C><><EFBFBD>!}2<07><>p1<70>v<0E>j$<24>.<2E>ᶀ<EFBFBD><E1B680>I<EFBFBD><<17><>`<60><07><10>Gy!<21>[y<>B<EFBFBD>O\<5C><><EFBFBD><EFBFBD><EFBFBD>܃<EFBFBD>s'<27>/v<>bl<><6C><EFBFBD><EFBFBD><EFBFBD>{|<7C>Z<EFBFBD>n<EFBFBD><6E><EFBFBD>?-<2D><>,<2C><>`<60>;<><7F><EFBFBD><EFBFBD><EFBFBD><EFBFBD>U<02>O~<7E> <09>?<3F><>_%<25><><EFBFBD>'<><7F><EFBFBD><EFBFBD><EFBFBD><EFBFBD>U<02>O~<7E><>ܟ<><DC9F><EFBFBD>ܟ<><DC9F>?p<>5<02>~MƱ<4D><10> SG<53>e<EFBFBD>-2<>+<2B>-<2D>1<EFBFBD>Qh<51><68><EFBFBD><EFBFBD><12>O<EFBFBD><4F>/<2F>6<EFBFBD>E<EFBFBD>ٞ<EFBFBD>} <1D>9//<2F>3&<26>m0?p<><70><EFBFBD>"={4@<40>;<3B><>?<3F><><EFBFBD>t<19><>E<EFBFBD>ٹ<EFBFBD><D9B9>/<2F><> <1F><><EFBFBD><EFBFBD><EFBFBD>P<EFBFBD><50><EFBFBD>á<EFBFBD>'6Շ<36> <1D>Y<EFBFBD><1B>~<7E><><EFBFBD>e<EFBFBD><65>m<EFBFBD><6D><EFBFBD>-<2D>ˊ<EFBFBD><CB8A><EFBFBD><EFBFBD><EFBFBD><EFBFBD>?<3F>Ԏ,B
|
||||
endstream
|
||||
endobj
|
||||
68 0 obj
|
||||
25389
|
||||
endobj
|
||||
69 0 obj
|
||||
[67 0 R]
|
||||
endobj
|
||||
70 0 obj
|
||||
<<
|
||||
/Resources 71 0 R
|
||||
/Type /Page
|
||||
/MediaBox [0 0 307 179]
|
||||
/CropBox [0 0 307 179]
|
||||
/BleedBox [0 0 307 179]
|
||||
/TrimBox [0 0 307 179]
|
||||
/Parent 72 0 R
|
||||
/Contents 69 0 R
|
||||
>>
|
||||
endobj
|
||||
73 0 obj
|
||||
<<
|
||||
/Type /FontDescriptor
|
||||
/FontName /EAAAAA+mwb_cmsy10
|
||||
/FontBBox [11 -215 942 727]
|
||||
/Flags 33
|
||||
/CapHeight 0
|
||||
/Ascent 727
|
||||
/Descent -215
|
||||
/ItalicAngle 0
|
||||
/StemV 0
|
||||
/MissingWidth 500
|
||||
/FontFile2 74 0 R
|
||||
/CIDSet 75 0 R
|
||||
>>
|
||||
endobj
|
||||
74 0 obj
|
||||
<<
|
||||
/Length1 1992
|
||||
/Length 76 0 R
|
||||
/Filter /FlateDecode
|
||||
>>
|
||||
stream
|
||||
x<EFBFBD><EFBFBD>Umle <1C>=<3D>Ҏ<EFBFBD>{<7B><18>`3^wL<77>zs<D7AD><73>%<25><>[<5B><>"<22>AZC<5A><43><EFBFBD>[<5B>خu e<>AM<41>/C<><06><><0F>D%:_r<5F><07><>4!HBb <12><><03> <1F><> <04><><EFBFBD><EFBFBD>m<EFBFBD>
|
||||
<EFBFBD><EFBFBD>><3E><><EFBFBD><EFBFBD><EFBFBD><EFBFBD>s<EFBFBD><73>/ ^<02><><EFBFBD>:<3A>j<6A><7F><14> <1F><><EFBFBD>xiZ<69><5A> |<7C><>'<13>T<EFBFBD><54>~<7E>l<><6C><EFBFBD>Tvv<76><76><EFBFBD><EFBFBD>Sd<64><7F>g҆<67><D286><EFBFBD><EFBFBD>7<EFBFBD><02>0<EFBFBD><30><EFBFBD>$<24>\qm#<23>U<EFBFBD>7<EFBFBD>s<EFBFBD>3<EFBFBD> l<>x<EFBFBD><78><EFBFBD>l~<7E>b<EFBFBD>{<7B><>gU<67><55>g
|
||||
l;<3B>"<22>9<EFBFBD><39>I=g,<2C><><EFBFBD><EFBFBD><EFBFBD><EFBFBD> <00>Q!_<>^~u<><75>K_<4B>2
|
||||
y<EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>'<27>.z<h<>ft-/<01>_<EFBFBD> l <0B><>]<5D><>K<EFBFBD>gq<1A><><06>cE<>0<EFBFBD>Èa/vAÓ<41><C393>6<EFBFBD>ʪ<EFBFBD>E<EFBFBD>F<46>>W<><57>J<EFBFBD>Y<EFBFBD>N<EFBFBD>,<0E>1q<31><71><EFBFBD>P<14><><><7F><EFBFBD>o<EFBFBD>~e<><65>ev<65><76>2E<><16>c<11>Q<EFBFBD>w)<29><>8fg<18>3<18>3<EFBFBD><33> <1C><> <1C><>
|
||||
u<EFBFBD>k<EFBFBD>5<EFBFBD>U<EFBFBD><EFBFBD><EFBFBD><EFBFBD>uen<EFBFBD>(<28> <1C><>H<EFBFBD>",͍<><CD8D><EFBFBD><EFBFBD>!UYl<07><>3&<26>Mzmb<6D><62>'L<16><><EFBFBD>HU<48><55>3<EFBFBD>vUYy<59>1<13><><EFBFBD><EFBFBD>VM<56>Hi=i<><69>D%S˺<>\<08><><EFBFBD><EFBFBD>> <0B>5'<27><>- <20>J<EFBFBD><4A><EFBFBD>n <1D>M<EFBFBD><4D><EFBFBD><EFBFBD>L1<4C>D<EFBFBD>IO&<0E><><0E>^S<>|<7C>r<EFBFBD> ]O><3E>Rl<52><6C><EFBFBD><EFBFBD><EFBFBD><CDB9>-<2D><>6<EFBFBD><36><EFBFBD>W<EFBFBD>l0<6C><05>ȹA<C8B9><0F><>сQMU%<25><><EFBFBD>EƉ<45>RX<0E>ɤ0Yk4+<2B><>R<EFBFBD><52><EFBFBD><EFBFBD><19>%<25>n<EFBFBD><6E><EFBFBD><EFBFBD><05>H<EFBFBD><48>+<2B>0EY<45>L<>2&'k<><6B><EFBFBD>QU<51><55><EFBFBD><17><>E<EFBFBD><45>X$-p<><70>&<26>L<EFBFBD>"<22>a<EFBFBD><61><EFBFBD>MX<4D>T<EFBFBD>Tao<61><6F>
|
||||
BU<02>rIU<49>̺vu<76><75>a<0F>7<EFBFBD>Y<EFBFBD>K<EFBFBD><4B>n<EFBFBD>-<2D>&<26><>#*<2A>ĸ9<C4B8> O<><4F>n<17><><EFBFBD><EFBFBD><07><><EFBFBD>y}<7D>-4<04> h"<22><>'\<5C><>Y<EFBFBD><18><><EFBFBD>x<EFBFBD>kC%B5<42>n<EFBFBD>ń`9<0F>U;<3B><>[ <1E>ֽ<EFBFBD>q_ .<2E>|<7C>[_<>ޮٰ<DEAE><D9B0><EFBFBD><EFBFBD>T]<5D>ɚ<EFBFBD><C99A>`<60>
|
||||
<EFBFBD><EFBFBD>U ;͞<>6<><36>c<EFBFBD><63>%Z!<21>_FΈg<CE88><67>h<EFBFBD><68><EFBFBD>y<EFBFBD>F'<27>ʌ<EFBFBD><CA8C><EFBFBD><18><><EFBFBD>
|
||||
<EFBFBD><EFBFBD><EFBFBD><EFBFBD>+<2B><><EFBFBD><EFBFBD>h<EFBFBD>R<EFBFBD>}B<>f<EFBFBD>p<>'<27><>ޥ.Kd
|
||||
<EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>k<07>m
|
||||
<EFBFBD><01><><EFBFBD>ѽw<D1BD><77>Bx<42><78>c[Ewq<77><17>r<EFBFBD>S<EFBFBD>a<EFBFBD><61><EFBFBD><EFBFBD>w0/;<3B>`<60>l>p<><70>Jv<4A><76>.<2E><>B<EFBFBD>LXG<58><05><><EFBFBD> uLr0<72>*<2A><>` -,<2C>`<60>l^s<><73>M<EFBFBD>t<EFBFBD><74><EFBFBD><EFBFBD>F <05>S<> <20>4<EFBFBD>io<69><6F><EFBFBD><EFBFBD>h<EFBFBD>:F<0F><>֠SO<53> KxzGI<47><49>;MJ<><4A><EFBFBD>$<24>El-Z<><5A> <0C><06>(<28>;I<><49><14>ˎ9D<39><44>Hu<><& <0F>*+<2B><>q<EFBFBD>dl<64>U<EFBFBD> ;<3B>A<1A><>ԡ<><D4A1>٩L*=-m
|
||||
<EFBFBD>I<EFBFBD><EFBFBD><EFBFBD>nE<EFBFBD><EFBFBD>b&+
|
||||
<05><><EFBFBD><EFBFBD>-<2D><>R@<40><>f<EFBFBD><66>eV<65>bFј*<19>@g<>KӇ<4B>Q]<5D><>'I<11>g<EFBFBD><67><EFBFBD>t&?<3F>H{<7B><><EFBFBD>U5J<35><4A>Q<EFBFBD><51>#TG<54>*<2A><><EFBFBD>;Ixtld<W<>
|
||||
<12><16><>*I<>"f<><66>du<02>ˉ<EFBFBD>4e/<><7F><EFBFBD>Xy<58><79>ɔ1<C994>s5<73><EFBFBD> Ҫ<><D2AA><EFBFBD><EFBFBD>O<EFBFBD>ޞ<EFBFBD><DE9E>P<EFBFBD><50><EFBFBD>` <20><11>R<EFBFBD><52>KjO<6A><4F>e<>"{,?o<><6F><EFBFBD><EFBFBD>Ek[<5B><><EFBFBD><06>9<EFBFBD><39>cі <1F>n<EFBFBD><EFBFBD><DC8F><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>}<7D>
|
||||
endstream
|
||||
endobj
|
||||
76 0 obj
|
||||
1324
|
||||
endobj
|
||||
75 0 obj
|
||||
<< /Length 77 0 R /Filter /FlateDecode >>
|
||||
stream
|
||||
x<EFBFBD>k <00> <00>
|
||||
endstream
|
||||
endobj
|
||||
77 0 obj
|
||||
9
|
||||
endobj
|
||||
78 0 obj
|
||||
<<
|
||||
/Type /Font
|
||||
/Subtype /Type0
|
||||
/BaseFont /EAAAAA+mwb_cmsy10
|
||||
/Encoding /Identity-H
|
||||
/ToUnicode 79 0 R
|
||||
/DescendantFonts [80 0 R]
|
||||
>>
|
||||
endobj
|
||||
80 0 obj
|
||||
<< /Type /Font
|
||||
/BaseFont /EAAAAA+mwb_cmsy10
|
||||
/CIDToGIDMap /Identity
|
||||
/Subtype /CIDFontType2
|
||||
/CIDSystemInfo << /Registry (Adobe) /Ordering (UCS) /Supplement 0 >>
|
||||
/FontDescriptor 73 0 R
|
||||
/DW 0
|
||||
/W [ 0 [750 776 ] ]
|
||||
>>
|
||||
endobj
|
||||
79 0 obj
|
||||
<< /Length 81 0 R /Filter /FlateDecode >>
|
||||
stream
|
||||
x<EFBFBD>]P<>j<EFBFBD>0<10><>+<2B><> <1E>l<EFBFBD> <0C><><EFBFBD>C <1F><><07><><EFBFBD>ԒX<D492><07>}<7D><>q z<><7A><EFBFBD>0<EFBFBD>h<EFBFBD>[<5B>l<04>E^<5E><18>X<EFBFBD> g<><67>Bp<><70><EFBFBD>j<EFBFBD>V<EFBFBD> <1D><>&
|
||||
<EFBFBD><EFBFBD><EFBFBD>:G<>:g<<3C>n,<2C><><EFBFBD> <20><>g<EFBFBD><67><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>/<2F><><EFBFBD><EFBFBD>'i$<24>F8<46><38><EFBFBD><EFBFBD><EFBFBD><EFBFBD><10>pB<17><> :ͯh<CDAF>e<EFBFBD><65><13>`<60>s<EFBFBD><13><><EFBFBD><EFBFBD><EFBFBD><0F><>j<>ՖAy<41>s<EFBFBD>
|
||||
I<EFBFBD><11>K<EFBFBD><4B><EFBFBD><EFBFBD>[<5B>&<26>?<3F><>M5u<>t<EFBFBD>M<EFBFBD><4D>Q<EFBFBD>PY<50>kwVv<56><76>8"<22><>(<28><>ep<65> <1C>:<<3C>|<7C>*><3E><><EFBFBD>}
|
||||
endstream
|
||||
endobj
|
||||
81 0 obj
|
||||
234
|
||||
endobj
|
||||
82 0 obj
|
||||
<<
|
||||
/Type /FontDescriptor
|
||||
/FontName /EAAAAB+mwa_cmmi10
|
||||
/FontBBox [-34 -250 1047 750]
|
||||
/Flags 33
|
||||
/CapHeight 683
|
||||
/Ascent 750
|
||||
/Descent -250
|
||||
/ItalicAngle 0
|
||||
/StemV 0
|
||||
/MissingWidth 500
|
||||
/FontFile2 83 0 R
|
||||
/CIDSet 84 0 R
|
||||
>>
|
||||
endobj
|
||||
83 0 obj
|
||||
<<
|
||||
/Length1 2576
|
||||
/Length 85 0 R
|
||||
/Filter /FlateDecode
|
||||
>>
|
||||
stream
|
||||
x<EFBFBD><EFBFBD>VkL[<5B>~<7E>s<EFBFBD><05><05><>c<EFBFBD> <1C>@<08>c<EFBFBD>
|
||||
<EFBFBD>$4Xք6!id4V8c{<7B>1<EFBFBD><04>E<EFBFBD>ZEUP<>m<EFBFBD><6D>*<2A><>VʪVݦ<03><>*JQ<4A>*Z*<2A><><EFBFBD><EFBFBD><EFBFBD>E<EFBFBD><45>j<EFBFBD><6A>H<EFBFBD><48> ݴ^b<><62><EFBFBD>a<EFBFBD><61>v<EFBFBD><76>s<EFBFBD><73>}<7D><><EFBFBD><EFBFBD><EFBFBD>w<EFBFBD> T<>Q<EFBFBD><51>=t<>-R<>@<40> <20>{lb~N<><4E><EFBFBD><EFBFBD>;<3B><>X<><58>L:Ϲ<CFB9><11> <1D>K<EFBFBD>NO<4E><4F><EFBFBD><EFBFBD>,<2C><><EFBFBD><EFBFBD>ɤ<EFBFBD><C9A4>O<EFBFBD><4F>8J<38>?"<22><>"<22><>n<>Lz4d<34>s<EFBFBD><73><<3C>%<25>C#k<>0A<30><41>Q<EFBFBD><51>F<EFBFBD>+<2B><><EFBFBD><EFBFBD>_<EFBFBD><1A>?#Y<><59><EFBFBD><EFBFBD>[ߌ<><10><>S<EFBFBD><53>f
|
||||
Ź<EFBFBD>s<EFBFBD> <00><1B>~f65s<35><73>I<><49>䝐`<60>ۡ<><DBA1><EFBFBD>h<EFBFBD><68><EFBFBD>E<><45>/<2F> <0B>e<EFBFBD>b<02><>M<EFBFBD><4D>E<EFBFBD><04>p?<3F><><EFBFBD><01><><EFBFBD><EFBFBD>.<2E><><EFBFBD>$F1<46><07> <00>#<23><><EFBFBD>]<5D>Dl<><6C>- Y|<7C>;<3B>u<EFBFBD><75><EFBFBD>O<EFBFBD><4F><EFBFBD>[<5B><>xS<78>"<22>R|L<#<23><0F><0E><><EFBFBD> <1F> <0B>8_<38>?<3F><><EFBFBD><EFBFBD><EFBFBD>W<EFBFBD>%<25>ۼ<EFBFBD>[<5B>ڸ&<26>ʞf{<7B>6ܦ*<2A>Du<44><75><EFBFBD><EFBFBD>:ޠz<DEA0><7A>*^<5E><15><><17>U<>.<2E><>r<EFBFBD>ϚU>EU.<2E><02><>Z<EFBFBD><5A>1<EFBFBD><31><19><>Iq¬| <1C>P<EFBFBD><50>p<EFBFBD>* D?UY@tǶx=5<>n<EFBFBD><6E>QUYa<59>Y-<2D>(<28> Cp<43><70>{B^<<3C><>ħ<EFBFBD><C4A7><EFBFBD><EFBFBD> q<><71><EFBFBD><EFBFBD>dm<64><6D>9<EFBFBD><39><EFBFBD>no<19>Y<EFBFBD><59>_<EFBFBD>l<><6C>ε<EFBFBD><CEB5><EFBFBD>;<3B>C<EFBFBD>/3<>1.<0F>mR<6D>i<>u!<21>h<EFBFBD><68>r,<2C>x4<19><><EFBFBD>@<40>ȡ<EFBFBD><01>Fy0<79>[4 <1D><><01>.&<26><1B>k<EFBFBD>#G<>J<EFBFBD><4A><EFBFBD><EFBFBD>M<EFBFBD>M<EFBFBD>[<5B><<3C>6<EFBFBD><36><EFBFBD>bc<62><63><EFBFBD><17>{?6)wum<75><1A>-<1B><>`<14>8<EFBFBD><38> -;<3B>? W<><03><>+<2B>Ɖ<><C689><EFBFBD><EFBFBD>XTH<54> J<>N<EFBFBD>0<EFBFBD>J<EFBFBD><4A><06>zL<7A>ͺ<EFBFBD>>?a<03><>~}<7D>Ȉ.*q
|
||||
<EFBFBD><EFBFBD>xV甸^<5E>R<EFBFBD><06><>L<EFBFBD><1A>
|
||||
7ĵ<EFBFBD><EFBFBD>TĊ[<5B>1<EFBFBD><31>2<EFBFBD>/<2F><15>i<EFBFBD>AjX<6A>hR<68><52><EFBFBD><18><> <20><12><04><1B> <0C><><EFBFBD>j<EFBFBD>W<EFBFBD><57><EFBFBD>X Iz<16><><EFBFBD>c ]{<7B>Ň<EFBFBD>i<EFBFBD> <20><><EFBFBD>ֳb<11><>-<2D>,<2C><><EFBFBD><EFBFBD><EFBFBD> A,<2C>-<1A><>EZ<45><5A><EFBFBD><EFBFBD>H8<48><0E><03>@?'<27>س<><D8B3><EFBFBD><EFBFBD>'/<2F> oSFw,.qc<><17>6(G 1<><31><EFBFBD><EFBFBD><EFBFBD><EFBFBD>J<EFBFBD>F<EFBFBD>ǎ ~<7E><1B><><EFBFBD><EFBFBD><EFBFBD>ܵS<DCB5><11><><EFBFBD>qV٭[<5B><11>ILs<4C><73>a<EFBFBD><61>hm<68><6D><EFBFBD><EFBFBD><EFBFBD>E#]<5D><><EFBFBD><EFBFBD>Ύ&2<><32>Ǣ<EFBFBD><C7A2><EFBFBD><EFBFBD>);,<2C><><0F>ܓg<DC93><67>S<EFBFBD>%J2S<02>1j<02><EFBFBD><7F> &UXφ<58><CF86><EFBFBD>'<27>v<EFBFBD>=k
|
||||
<EFBFBD><EFBFBD>e<EFBFBD><EFBFBD>n<EFBFBD>g<EFBFBD><EFBFBD>Y<EFBFBD>ה<EFBFBD>Z<EFBFBD> <1E>iӶ<69><D3B6>IJ<49><EFBFBD>
|
||||
<EFBFBD>:y<><79><EFBFBD>%<25><>D!l5~<7E><>t<EFBFBD>BB<42>c<EFBFBD><63>[<5B>[<5B><>s9<73><39>R
|
||||
<EFBFBD>ǿ)æ<><C3A6>K <1E><><EFBFBD>:<3A><><EFBFBD>p<EFBFBD><13>t<EFBFBD><74>O0(u<7F><75>b3<62>1<EFBFBD><31>X0<14>ܒ<EFBFBD><DC92><EFBFBD>j+9<>N&5<> NNhU䯭7<E4AFAD> qLdQ<64><51><EFBFBD>֯<EFBFBD> <1E><><EFBFBD>^<16><>uQGx<47><78> <1D>V5ou=<13>T<EFBFBD>><3E>
|
||||
<EFBFBD>>c<7F><63>Y<EFBFBD><59>»<EFBFBD><C2BB>No<4E><6F><16><><EFBFBD><EFBFBD><EFBFBD>z<EFBFBD><7A><EFBFBD><EFBFBD><EFBFBD><EFBFBD>ˁ<EFBFBD>
|
||||
&<26>¾<0F><>Ev(<28><><EFBFBD><EFBFBD>3<EFBFBD><33>h<EFBFBD>\<5C><><EFBFBD>1z<31><7A>/<2F><><EFBFBD>/ѭk<D1AD>g<EFBFBD><67><EFBFBD>ǹ<EFBFBD>.ۘ<><DB98><EFBFBD><EFBFBD><EFBFBD>5<EFBFBD>7ӽ/o<>kWK<57> <1E> <0C>J<EFBFBD>lÏ<6C><C38F>7(ҟ<> ߌ<>1x<31><0E><>pP<A<><41><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD> <1C><><12><10>Fo<46><6F><EFBFBD><EFBFBD>w<EFBFBD>F<EFBFBD>1<EFBFBD><31><EFBFBD><EFBFBD><EFBFBD>v<EFBFBD>1 ;\<5C><<3C><>l d<><64>2Q<><51>*c<><63>ɒ 6<>VM/3x<33>\<5C> <1C><><EFBFBD>y<EFBFBD>`#e,<2C><>Se,b+<2B><>X"<22>H<><48><19>f<>E<19>A<EFBFBD>.b<>i<0E><>~ <12>C<EFBFBD><43>#<23> <1C><08>y<EFBFBD><79> y<16>+<2B>y<EFBFBD><79>#ưތV4<56><14>)<29>1O<31>I<EFBFBD>l<EFBFBD><6C>3<><01>oD<6F>i*`<60><>aTeDJQ<4A><51>(<28><>1<EFBFBD>8hfJ<66>F%<25><10><<3C><>ӳ<EFBFBD>tfNޕh<DE95><68><EFBFBD><EFBFBD> A9<41><15>9<EFBFBD>@H <1E>r<EFBFBD><72><EFBFBD><EFBFBD>|H<><48><EFBFBD><EFBFBD>aì(<0F><><EFBFBD><EFBFBD><EFBFBD><EFBFBD>d<EFBFBD>= <1C>k
|
||||
C<EFBFBD><X<>&E<><45>˥&沅<><E6B285>|05<30>ʩqJ<71><4A>U7Fu<46>闥<EFBFBD>ۉ\<5C><>&<26><>l<EFBFBD><6C><EFBFBD><EFBFBD>4m<34>92<39>%1<>><3E><>l,gȜ<67><C89C><EFBFBD><EFBFBD>|><3E><>w<EFBFBD>0<EFBFBD>N<EFBFBD><4E>و<EFBFBD>?Ӆi<D385>a<EFBFBD><61><EFBFBD>S黣@<40>Pͫ<50>p(<28><16><><16><><EFBFBD><EFBFBD><EFBFBD>#<23>y<17>c<EFBFBD><63><EFBFBD>}t<><74><EFBFBD><EFBFBD><EFBFBD><EFBFBD>7<EFBFBD>W<>МtT9<54>pOWd<57>I!<21>T<EFBFBD>:<3A><><EFBFBD>|<7C><><EFBFBD>Z[<5B>4<><34>?
|
||||
endstream
|
||||
endobj
|
||||
85 0 obj
|
||||
1742
|
||||
endobj
|
||||
84 0 obj
|
||||
<< /Length 86 0 R /Filter /FlateDecode >>
|
||||
stream
|
||||
x<EFBFBD>k` % ( <00>
|
||||
endstream
|
||||
endobj
|
||||
86 0 obj
|
||||
12
|
||||
endobj
|
||||
87 0 obj
|
||||
<<
|
||||
/Type /Font
|
||||
/Subtype /Type0
|
||||
/BaseFont /EAAAAB+mwa_cmmi10
|
||||
/Encoding /Identity-H
|
||||
/ToUnicode 88 0 R
|
||||
/DescendantFonts [89 0 R]
|
||||
>>
|
||||
endobj
|
||||
89 0 obj
|
||||
<< /Type /Font
|
||||
/BaseFont /EAAAAB+mwa_cmmi10
|
||||
/CIDToGIDMap /Identity
|
||||
/Subtype /CIDFontType2
|
||||
/CIDSystemInfo << /Registry (Adobe) /Ordering (UCS) /Supplement 0 >>
|
||||
/FontDescriptor 82 0 R
|
||||
/DW 0
|
||||
/W [ 0 [365 831 680 ] ]
|
||||
>>
|
||||
endobj
|
||||
88 0 obj
|
||||
<< /Length 90 0 R /Filter /FlateDecode >>
|
||||
stream
|
||||
x<EFBFBD>]P<>j<EFBFBD>0 <0C><>+t<>=,N<>=<3D><>JJ!<21>mK<6D>~<7E>c˩<63><CBA9><EFBFBD><EFBFBD> <1C><><EFBFBD><EFBFBD><EFBFBD>B~<7E>43 m<><6D>9A<><41>W=F0<46>i<EFBFBD><69>/<2F> <1C>+<2B><1A>UqG|<7C>I<EFBFBD>B$q<><71><11><><><D78D><EFBFBD>3<01>G<EFBFBD>̑V8=i?<3F>4<><34>#<23>d<EFBFBD><08><><EFBFBD>?<3F><><12>N<>"<22><>C<EFBFBD><43><15>]<5D>W9!<08><>t:<3A>l\/I<><49><EFBFBD>\B<42>2(<28>qR!I7bq+S5p{I<>d<EFBFBD><EFBFBD><7F><EFBFBD><1A><><EFBFBD>t<EFBFBD>M<EFBFBD><4D>Q<EFBFBD>PY><n<>ސb<DE90>]<5D>=<3D>n<EFBFBD><6E>j!JYy5 2dz<0E> <1F><><EFBFBD>/<2F><>
|
||||
endstream
|
||||
endobj
|
||||
90 0 obj
|
||||
241
|
||||
endobj
|
||||
91 0 obj
|
||||
<<
|
||||
/Type /FontDescriptor
|
||||
/FontName /EAAAAC+mwa_cmr10
|
||||
/FontBBox [-43 -250 1008 750]
|
||||
/Flags 33
|
||||
/CapHeight 683
|
||||
/Ascent 750
|
||||
/Descent -250
|
||||
/ItalicAngle 0
|
||||
/StemV 0
|
||||
/MissingWidth 500
|
||||
/FontFile2 92 0 R
|
||||
/CIDSet 93 0 R
|
||||
>>
|
||||
endobj
|
||||
92 0 obj
|
||||
<<
|
||||
/Length1 6620
|
||||
/Length 94 0 R
|
||||
/Filter /FlateDecode
|
||||
>>
|
||||
stream
|
||||
x<EFBFBD><EFBFBD>Y p<>ו><3E>I<7F>$˖<><CB96>e<EFBFBD><65><EFBFBD>%<25><><EFBFBD>Y<12><><EFBFBD>e<EFBFBD>#l<> <0B>?<3F><><EFBFBD>`[<5B><02>B<EFBFBD><42>ii<69>dؔI<D894><49>R<EFBFBD><52>S<EFBFBD><53>a<EFBFBD>m<>n<EFBFBD>%n<>d<EFBFBD>6<EFBFBD><36>ng<6E>Ͷ<EFBFBD>L<EFBFBD><4C><EFBFBD>X<EFBFBD><58>Ʉ4mW<6D><57>{<7B><><EFBFBD><EFBFBD><EFBFBD>s<EFBFBD>s<EFBFBD><73><EFBFBD>@ @<01><><EFBFBD><EFBFBD><EFBFBD><EFBFBD>e<EFBFBD><65><1B><> <00><17>w<EFBFBD><77><EFBFBD> S<>d<EFBFBD>Y<EFBFBD><59><05>o<EFBFBD>ώMS<4D> <02>KX<4B><1A><>o<EFBFBD>x<EFBFBD><78> <00>C <00>K㣃# <>'1 <00><02><>DZ#c<1B><>v<04>E<EFBFBD><45>{'ޖƱ=<3D><><EFBFBD><EFBFBD><EFBFBD>a\<5C><>*|<7C><18>g<EFBFBD>g 7q<37><17>6<EFBFBD> <1C> }o]<5D>۸><3E><><EFBFBD><EFBFBD>ܞ{<7B>@`<60><>0><3E>{t<><0F><>¶<>o1p %<25><><13><><EFBFBD><EFBFBD>ܽ%\y<>><3E>.<2E><17>k<EFBFBD> <1E>o@.<2E><>p<01><><EFBFBD>p <1E>i<EFBFBD><69>q<EFBFBD>
|
||||
<EFBFBD> <06><><06><10>:B |<7C><>20K<30><4B>߳<EFBFBD><DFB3>7<EFBFBD>{<7B><><EFBFBD><EFBFBD>X)<29>.#<23><><EFBFBD> <1D>@}B}H<><1B>5M<35>R<0E><><EFBFBD>!/<2F><><EFBFBD>!2Ml<4D>{<7B><>]<5D><>m<EFBFBD><6D>gp n<><6E>p ~ <00><>E<EFBFBD>
|
||||
<EFBFBD>E-<2D><><EFBFBD><EFBFBD>ex <1E><>p<02>9q<><71><19>i<08>t <09> f<><66>*<2A><>N<EFBFBD><4E>V<EFBFBD><56>2<EFBFBD>
|
||||
<EFBFBD><EFBFBD><EFBFBD>J8<EFBFBD><EFBFBD>)<02><>xM<7F>SC<53><43><EFBFBD><EFBFBD><EFBFBD>ܮ<EFBFBD><DCAE> M<><4D> pv<70><76>{ĪV<C4AA><12>Ҟi<><69><EFBFBD>.<2E>ˉ <09><><EFBFBD>J}<7D><>X<>Wm <0B>ZL<5A><4C># چ<>)<29>4YM#WB<57>S<EFBFBD><53>f<EFBFBD>\s^OG<4F> 06S<36>xBbk L<><4C>n<EFBFBD>1<EFBFBD>6<EFBFBD><36>1<0E><>l춖Y?6<>Y<EFBFBD><59><EFBFBD>ο<EFBFBD><CEBF><EFBFBD><EFBFBD>ߜ<EFBFBD><DF9C><EFBFBD>
|
||||
w><3E>p<EFBFBD><70>b<EFBFBD>J<EFBFBD><4A>v<EFBFBD>^<5E>cA+J<>jN<6A>M<EFBFBD><03><><EFBFBD><EFBFBD><EFBFBD>lJ<6C><4A><EFBFBD>UB<55>M<EFBFBD>M֦SV H0<48>pNB.<2E>W<>i<EFBFBD>H<EFBFBD>M<>މF#ʂ<>h1&<26>== <09><1A>%<25>5<<3C><><EFBFBD>ᄢ4<E184A2><34>25<32>Gܮ <1F>
|
||||
<EFBFBD>{/<2F>Q<EFBFBD><51>*<16><>d<EFBFBD>Bѧ<42><D1A7> <0C><><EFBFBD>B<EFBFBD>݉<EFBFBD>@<40><><EFBFBD>ۅCM<43>&T<><02>.<2E><>u<EFBFBD><75>jHhJݐ<4A>tc<74><63>~E<> r<05><> 3 <0B><><EFBFBD><EFBFBD><05>-HP<48><50>?JOb<4F><62>G_<47><5F><EFBFBD>2<EFBFBD>1<EFBFBD><31> <00><05>paA U<><02>d <00><>g<EFBFBD><67>y><3E>z(<28><><EFBFBD>ۅ^/<01>e<>W%<25><>W^<5E><>_<EFBFBD>^<5E>)<14>
|
||||
-t<>B<EFBFBD>U <09><><EFBFBD>Z<EFBFBD>!<21>^<5E>Ye3<65><33>ʔ,"<22><><EFBFBD>lן<6C>a^C<08><1A><>l'<1A>(<28>/# <0C>H8G<18>UB<>yқ<79><10>dg<64>Ng<4E><67><EFBFBD><EFBFBD>3<EFBFBD>T<EFBFBD><54>W<19>V<EFBFBD><56>*<2A>Y{<7B><><EFBFBD>sw<0F>K2D<19>7zT.<2E><><EFBFBD>,DAǽ%<25>2<EFBFBD><32> &<26><><EFBFBD>r H<> <1C><><EFBFBD>b<EFBFBD>`<60><>h<EFBFBD>( nP<01>r<EFBFBD>B<EFBFBD><42><EFBFBD>~oe<01><>P<EFBFBD><50>ү<EFBFBD>UQv<51><76><EFBFBD>vE<><45><EFBFBD>6<EFBFBD>|<7C>ĥ^s<><73><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>3<>+/<2F><>&ϐ<><CF90><EFBFBD><EFBFBD><EFBFBD>c<EFBFBD><63>p4<70>|<7C>M<EFBFBD><EFBFBD><7F><EFBFBD>Of<><17><><EFBFBD>w<EFBFBD><77><EFBFBD><EFBFBD><EFBFBD>y7<79>>Q<><51><EFBFBD>j<><11>:<02><><EFBFBD><EFBFBD><EFBFBD><<3C>`s)<18><><14>儣<EFBFBD>u<EFBFBD>"ͨU<CDA8><55>¯<EFBFBD><C2AF><EFBFBD><EFBFBD><EFBFBD><EFBFBD>
|
||||
<1D><>e<02><>RL<52><4C>h#pL<70>`Լ<><P@<40><>~_<><5F>rK <1D><>*뉯J<EB89AF>]RO{i8<69><38><EFBFBD><EFBFBD><1A><><EFBFBD>D<EFBFBD><44><EFBFBD>NM<>97ʟ۲0<DBB2><30><EFBFBD><EFBFBD>pR<70><52><15><><EFBFBD><EFBFBD><EFBFBD>~<7E><><EFBFBD>O}<7D><> <1D><>fg<66><67><EFBFBD>ֶG:z<> xuv<75><76><EFBFBD><11>9P<39>R<EFBFBD>lK<6C><4B>Gj<47><6A><EFBFBD>f<EFBFBD><66><EFBFBD><EFBFBD><EFBFBD><EFBFBD>
|
||||
<EFBFBD>r<>B!W<><57>M<EFBFBD><4D><EFBFBD><EFBFBD>MOL<4F><4C>F<EFBFBD><EFBFBD>{c<><63><EFBFBD><EFBFBD><EFBFBD><14>2UF<55>.<2E><07><>.<2E> =%9t<39><74>WgQڤ<><DAA4>ĩ A1%<25><><EFBFBD><15><><EFBFBD>C<EFBFBD>/<2F>y<EFBFBD>HJ<48>h:<3A>%4GU<47>m!<21><>*@<40>i4<11><><EFBFBD><7F>T<EFBFBD><54>J*S<>W<EFBFBD> <1F><><EFBFBD>R <1F><><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>s6_<36><5F>46z<36>k<> m<><6D><EFBFBD>|<7C><><EFBFBD><05>q<EFBFBD><71>\<5C><><EFBFBD>&<26>rd<72>}<7D><>c <1D>_<EFBFBD><5F><EFBFBD>E<EFBFBD>W<EFBFBD>56<35><36>NL<4E><1A>n<EFBFBD><6E><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>e<EFBFBD><65>o$<24><>35<33>/<2F><>.<2E> <1D><>J#<23>t|<16>,R%<25><11>Β<EFBFBD>QL><3E><><EFBFBD>s;<3B>'<27>y<EFBFBD> k%<1B><><EFBFBD><EFBFBD><EFBFBD>E<EFBFBD>{ģT<C4A3>TUV<55>9<EFBFBD><39><EFBFBD><EFBFBD>f<EFBFBD>D<EFBFBD><44>
|
||||
7<EFBFBD>UY<EFBFBD>xN"xA<78><41>3Q T<><54>~<7E><>?<3F>Jr <1F><18><05><><EFBFBD><EFBFBD><EFBFBD> <09>^1<>M:<3A><<3C>e<EFBFBD><65> <0B><><EFBFBD><EFBFBD><EFBFBD>o<EFBFBD><6F><EFBFBD>?
|
||||
^B0_ u<>~<14><> <09><>D<EFBFBD><44><EFBFBD><EFBFBD>L<EFBFBD>BJ<18> <0C><> 5<><35><19><>d<EFBFBD><64>ղ<EFBFBD>&<26>&z#<23>M%<25><>g<EFBFBD>~b<01><><EFBFBD>ƫ<EFBFBD><C6AB><EFBFBD><EFBFBD><EFBFBD><1B>ͼos݀<73>a<EFBFBD>;'\?<3F><><EFBFBD>TY<54>B <1E><><EFBFBD>g&<1A><05><><EFBFBD>~<7E><><EFBFBD>dBЉ<42><D089>{w<><18>e<EFBFBD><65>b<EFBFBD><62><EFBFBD>yT<79><54><EFBFBD>ۥ2<DBA5>[2<><32><14><> <19><><EFBFBD>"eͤSŠAKSn,<2C><> <00><15>[<5B>Z<EFBFBD>N0Z<30><5A>I<>VI<56>9
|
||||
<EFBFBD><x{a<><61><EFBFBD><EFBFBD>ӽ <1E><>w <1E>ZÎ<5A>JJ.Qk<51>_J<5F>ɒsdÞ<64><C39E>G<EFBFBD>ٷ<EFBFBD><D9B7><EFBFBD><EFBFBD><EFBFBD>s}F<> <1F><1A><><EFBFBD>s<EFBFBD><EY۞<59>x;g"<22>9&<26>64"<22><>
|
||||
<EFBFBD>TDC<EFBFBD>6 <1A>%yv)<29><><EFBFBD><05><>b<>j<EFBFBD>> <1D>床30<04>іD<D196><44><EFBFBD><EFBFBD>TJq<4A>><3E>U<EFBFBD><55>Z<EFBFBD>(U <09>*<2A>D<><44>-<2D><>Ot<4F>?<3F><>q<EFBFBD>v|<7C><>\<5C><><EFBFBD><EFBFBD>˷<EFBFBD> <1E>h<EFBFBD><68><EFBFBD><EFBFBD>\]T<><54><EFBFBD>k<EFBFBD>)%<25>C<EFBFBD><43> <1E>V<EFBFBD><56>7
|
||||
<EFBFBD>^<5E>;W<><57>J.<2E><><EFBFBD><EFBFBD><EFBFBD><17> <1C>b(<28><>#x<>u<EFBFBD><75>0<EFBFBD>Ļ<EFBFBD> 3Q<33><51>y<EFBFBD><79>0y<30>N&<26><>-"<22>O<EFBFBD><4F>E<EFBFBD><45>I<EFBFBD><49>(<28>h<EFBFBD><06><><03><>WiR<69><52><EFBFBD><EFBFBD><EFBFBD><EFBFBD> <09><><EFBFBD>Z<EFBFBD>}<7D>~ml<6D>Ö<EFBFBD>F<EFBFBD><46>f<EFBFBD><66><EFBFBD>d<EFBFBD><64>oHn<48> X<><58><EFBFBD>㺶<EFBFBD>L<EFBFBD>Z~<7E><><EFBFBD>n8(`<60>}HЯ|!<21><>(<28>)
|
||||
<EFBFBD> ^ <00><>H<>:_<><16> <00>W<EFBFBD><57>M<EFBFBD><4D><EFBFBD>o@T<><54>9<EFBFBD>/<04>TG]
|
||||
<EFBFBD>%%<25><><EFBFBD><EFBFBD>c<EFBFBD>-IgKoM<6F><4D>W{<7B> @j<><6A><EFBFBD>/$5N]<14><><EFBFBD>F<EFBFBD>\<5C>'<27><>%<25>I<><49><14>C<EFBFBD><43>E%<25>+<2B>uE^ <0B>R٦1<>.W<>خ\o<><6F>y<EFBFBD><79>w <>.:Pʮ<50>*Ĭ.<2E> j<>^<5E>3<EFBFBD><33><03><><EFBFBD>KJ<4B>=<3D><><EFBFBD>C<EFBFBD><43><EFBFBD><EFBFBD>Ã'
|
||||
YNVxM<4D><7F>x<EFBFBD>7Ƿ<37>(<06><>|<12><>/m <1D>~|a@!<21>pq<70><71><EFBFBD><EFBFBD><EFBFBD>d?<3F>E<EFBFBD>!W<><57><EFBFBD>(<28>Ux<>%<25>\<5C>\f4h,<2C>V<EFBFBD>5<EFBFBD>t<EFBFBD><74><EFBFBD>,<2C>Hy<48><79><EFBFBD>,<2C>F j D0<44>Yk<59>j<EFBFBD>F<EFBFBD>C<><43><EFBFBD>$DI<44><49>O'<27>3<EFBFBD><33>#<23><><EFBFBD>k<EFBFBD><6B><EFBFBD><EFBFBD><EFBFBD>.귯qF<71><46>t6l<36>M><3E><>v*<2A>vW<76>=<3D>D<EFBFBD>J&
|
||||
o\<5C><>Ǯ<EFBFBD>F+<2B><><EFBFBD><EFBFBD><EFBFBD>GY<47><06>&<1A><><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>*}<7D><>\<5C><><EFBFBD>J<>!<21>S<>PHa<48><61>BVR<><52><EFBFBD><EFBFBD>K<>X§Sy<53>ʬ<EFBFBD><CAAC><EFBFBD><EFBFBD>HΏ~<7E>D<EFBFBD> <1F><>B Rq<52><71><EFBFBD>q<EFBFBD>
|
||||
5<EFBFBD>2KF! <20><><EFBFBD><EFBFBD>
|
||||
<EFBFBD>xC6<43>Bm5<6D>xF<78><46> <0B><06><>bMd<4D>B<08>b<EFBFBD>C[<5B><><EFBFBD>
|
||||
<EFBFBD>h$8+<2B>F<EFBFBD><46>{P¨L<C2A8>.Q<>K<EFBFBD><4B><EFBFBD><EFBFBD>/}i<><69>R*G<1B>J<EFBFBD><4A>P<EFBFBD><50><EFBFBD><EFBFBD>M9<4D><39>H<EFBFBD><48>j̃><3E><>B1<42>eP=<3D>%{<7B><>`$<24>ۀ'<27><>|<7C>6)=<3D><><EFBFBD>4<>#<23>
|
||||
<EFBFBD><EFBFBD>B<EFBFBD><0F>"<22>5<EFBFBD>,<2C>F#!<21><><EFBFBD>Kw<4B>9<EFBFBD><39><EFBFBD><EFBFBD>]<5D><>A=u<><75>`<60>o<EFBFBD><6F>Y(u<>K<EFBFBD>V}<7D><><EFBFBD><EFBFBD><EFBFBD><EFBFBD> [I,C<>ट|<7C>+u L<>E̸<45>@Z<><13>ND<4E><44><EFBFBD><1A>@<40><>݅<EFBFBD><DD85>3<EFBFBD><33>?<3F>11$!&<26><>}$<24><><07>(<28><>6<08><>f<EFBFBD><66><EFBFBD>.<2E><>|<7C>hw]<5D>`w[<5B>\<5C>-<2D><>l#<23><16>M <1C>m<EFBFBD>P<EFBFBD><50><EFBFBD>+<2B><<3C>{?cq<63><71><EFBFBD> B3<42><33>D<EFBFBD><1A><>:<3A>.5<EFBFBD><EFBFBD>KK˗<EFBFBD><EFBFBD>Rސ<EFBFBD>'<27>K<EFBFBD>8<EFBFBD>n<EFBFBD><6E><EFBFBD>ƙ<13><15>3<17><>;<3B><><EFBFBD><EFBFBD>= <1F>b,<2C><>/<><7F>\ <09><><1A><><13><>y<EFBFBD> ]YE<59><45><EFBFBD> j8W!巗aX<61><58>~<7E>f6H\<5C><><EFBFBD>Ҝ<EFBFBD>x<EFBFBD><78>7<EFBFBD><37><EFBFBD>ѡT T<><54><EFBFBD>):"y<><79>EO<11>O<EFBFBD><4F>c<><63><EFBFBD><EFBFBD><EFBFBD># <0B>RV<52><56><EFBFBD>c/<2F><>x<EFBFBD>P<EFBFBD>d<EFBFBD><64>[S\Rss~<7E><> <1D>"<22><><EFBFBD>|CY<43>f<EFBFBD><66><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>>?
|
||||
O<EFBFBD><EFBFBD><EFBFBD>8<EFBFBD>ʾ<EFBFBD>6tgZ낎<5A><EB828E><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>c<EFBFBD>IoqS<71>q<EFBFBD><71>gcŖP[<5B>gx}t(<28><>፺<EFBFBD>rP<72><50><EFBFBD>ߣ<EFBFBD><DFA3><EFBFBD><EFBFBD>Z<EFBFBD> }<7D><>YJmX<6D>kH<6B>
|
||||
<EFBFBD>R7 <00><>Rb_W<5F><57>Ѥ<EFBFBD>E <1C><><EFBFBD>0r0L)i<><69>@\<5C>O
|
||||
WET<EFBFBD>YE<EFBFBD><14><13><>%<25>G
|
||||
ׂ<EFBFBD><EFBFBD>{<7B>S3=F7!<21><05>\L<>Y<EFBFBD>`<60><><EFBFBD> -<2D><>0ic<69>[i <11><>HjE"D*S<>V<EFBFBD><56>Փ<EFBFBD>K<EFBFBD>u[C<>
|
||||
<EFBFBD>K<><4B>]<5D><><EFBFBD>\<5C>ar<61><72>=<3D><><EFBFBD><EFBFBD>-<2D><><EFBFBD><EFBFBD><EFBFBD>n ?<3F><><EFBFBD><EFBFBD><EFBFBD>ss<73>϶4?;s<> <17>֗<13><>5##k<>֝_=<3D>/<2F><>Xӿ<10><>9[<5B>
|
||||
.<2E><>n<EFBFBD>g<EFBFBD><67><EFBFBD><EFBFBD>m><3E>VW\<5C><><EFBFBD>:<16>]<5D><>z<><7A><EFBFBD><EFBFBD>[<5B>w4<77>7<EFBFBD><37>><3E>q<EFBFBD><71><EFBFBD><EFBFBD><EFBFBD><16> `|!I<><17>9<EFBFBD><39><16><><EFBFBD><EFBFBD><EFBFBD>w<EFBFBD><14><>gRY<<3C><>e<EFBFBD><65>%<05>4u?K<><4B><EFBFBD>)d<>Oɸ<4F><C9B8><EFBFBD><EFBFBD>[<5B>K <1C><><EFBFBD>sW<><57> _$#bRZ_<5A><5F>01wܑ<>+I<><49>{w<><77><EFBFBD><EFBFBD><EFBFBD>E!<21><><EFBFBD>L^<5E><04>X<EFBFBD>2Z<32>Yd<10><>DF&<26>j<EFBFBD>*pD<70><44> <1C>/<2F>#p<> :0j<30>onZ<6E>_<EFBFBD><5F><EFBFBD> <0B>
|
||||
s<><73><EFBFBD>5-<2D><><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>D<EFBFBD>|~"y"<22><><EFBFBD><EFBFBD><EFBFBD>'<27><>C;<3B>L<EFBFBD>"<22>A<> Z<>u!<21><><14>x9)vHe<48>4s<><73>n@<40>Aa:W+r<><72>_
|
||||
L<EFBFBD>=<3D><><12><>|<7C>D<EFBFBD>I<EFBFBD><49><EFBFBD>
|
||||
[*Q<><T*o<>ȇ<EFBFBD><C887><EFBFBD><EFBFBD>~p<>#f<> e<><65><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>o<EFBFBD>{<7B><><EFBFBD>ZsV<73>yw<\<5C><> <1F><><EFBFBD><EFBFBD>75<>"/<2F>)<29><>[d G<>"<22>4u <0C><><EFBFBD>r<EFBFBD><72><EFBFBD><EFBFBD>4<16>'<27>u<0E>ݒ<EFBFBD>X
|
||||
͡Z<EFBFBD>#[Ͱe2><3E><><EFBFBD>2<EFBFBD>I<EFBFBD><49>FH<46><48>,!<21><^<5E><>UJ<55>"<22><>VIX/ёI Qs<51><73>O=A%&<26>45<15><>örF}<7D>/<2F><>4<EFBFBD>tq(<28>tz<74><7A>^q<08><><EFBFBD><EFBFBD>95<39><35><EFBFBD>_m]l<><6C>L<><4C><EFBFBD>7<EFBFBD><37>7<EFBFBD>Z<EFBFBD><5A><EFBFBD>*9<>*'<27>Z_4<5F><34><EFBFBD><EFBFBD><EFBFBD><EFBFBD>n" <0B>mwm<77><6D><EFBFBD><EFBFBD><EFBFBD>ŊM <1C>P<EFBFBD><50><EFBFBD>dɸW~ ;<3B>Z<EFBFBD>w<EFBFBD><77><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD> L<>><3E><>ҤM<D2A4><4D>:<3A><>h<EFBFBD>u_S<5F>c<EFBFBD>Ƭ<EFBFBD>;ODgn-<16>W<EFBFBD><57>Ļ<EFBFBD>!<21>u#7<>7ԇ<37><D487><EFBFBD>ԌH<D48C><48>R<EFBFBD>)Z1<>u"<22><>e <20>Rہ<52>'<05>'<27><><EFBFBD><19>|<7C>*<2A><>r<EFBFBD>?u <00><>ݏ<01>;Т<>S+os<6F>J<EFBFBD><4A>A<EFBFBD>V=<3D><19><>t<EFBFBD><74><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD> ^<5E>:<3A><><1A>o<EFBFBD>x<EFBFBD>D<EFBFBD><44><EFBFBD>;<3B><EFBFBD><1B><><EFBFBD>p<EFBFBD><70>œO#<03>:Wo[w<>L8<4C><38>G<EFBFBD>V<EFBFBD><56>
|
||||
<EFBFBD>K<EFBFBD><EFBFBD><EFBFBD>@9!<15>)<29>.<14> C <20>t> <茗"<22><A<7F><41><EFBFBD><EFBFBD><EFBFBD><EFBFBD> <0B>yn<79><6E>R<EFBFBD>+<2B>0gdA<64><41>;<3B><><EFBFBD> <0B>"<22><>:34<33>v<EFBFBD>@<40><><EFBFBD><EFBFBD> <20><>ʻ(qo<>v<EFBFBD><76>_N<5F>b~<14>}_<><5F><EFBFBD><0F>lq<6C><71>}v<>l<EFBFBD><6C><EFBFBD><EFBFBD><EFBFBD>9<EFBFBD>][ <1C> |<7C><><EFBFBD>ˏ<EFBFBD> <><7F><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD><EFBFBD>6v"`<60>k<>5<EFBFBD><07><>w ><3E>!<21>v<EFBFBD>E`Qx<51><78>H<><48>j<EFBFBD>9<EFBFBD>N<EFBFBD>Q}9 <>3 b/${<01>><3E><><EFBFBD><EFBFBD>uS<75><53>t<EFBFBD><74><EFBFBD>2k<32>
|
||||
<EFBFBD>b<EFBFBD>+c
|
||||
/<2F><><EFBFBD><EFBFBD>G<EFBFBD><47> <00> <0C>J?<3F>߸ FKi<4B><69><EFBFBD><EFBFBD><EFBFBD>J<EFBFBD><4A>W\<5C><18><>!<21> \<5C>ҁe<13><>l h3<68><33>3P<33>p|<17> <05>l;<3B> |<7C>;5<08>(<28>%<25><10><>bX<62>vm<07><17><><EFBFBD><EFBFBD>þK<4B><C298><06><>T<EFBFBD>R<EFBFBD>Au(݄<><DD84>>J<>|_<>m<EFBFBD><6D><EFBFBD>s<EFBFBD><73>*-S<><53><EFBFBD><EFBFBD>LCYL<59> <0C>y&-<2D><>An<41>e<0E><> g<12><><EFBFBD>u<EFBFBD>)A&<26>&<26><>L<EFBFBD><4C>ԧeLdSZfpΣi<CEA3><69>\<5C>ʹ<EFBFBD>a<EFBFBD>mh<6D><19><>8<EFBFBD><38> <18>q<EFBFBD>#<23>E7A <09>x<>j<EFBFBD><17>a<18>9<EFBFBD>3<EFBFBD><33>Z<EFBFBD><5A>&hÞ)<29><17>q<18>y<EFBFBD><79>{<7B><19><><EFBFBD>͉<EFBFBD>Q<EFBFBD>Gq<47>y<EFBFBD> <1E><><EFBFBD><EFBFBD>v<EFBFBD><76><EFBFBD>Z<EFBFBD>_X<5F>-(<28>L?!h%<25>4
|
||||
<EFBFBD><EFBFBD>3!<21>Z<>w<1A>7><3E>h<><68><EFBFBD><EFBFBD><EFBFBD><1B>c*n*1UTWW<57>L<EFBFBD><4C><EFBFBD><EFBFBD>)<29>Z<EFBFBD><5A>mp*>7>:<3A>15NM<4E>b´9Sltnt<6E><74>舧<EFBFBD><E888A7>2<<3C>v<EFBFBD>m<EFBFBD><6D>2<EFBFBD><07>f<EFBFBD><66>F<EFBFBD><46>L<EFBFBD><4C>t<EFBFBD>֍ΏN<CE8F>ø<EFBFBD>4,<2C>v<EFBFBD>P<EFBFBD>iԿ5<><35><EFBFBD><EFBFBD>m<EFBFBD>ӻ+PN6{Q<>A <1C><><EFBFBD><EFBFBD>ީAV<><56>&<26>c<EFBFBD><63>k \Cزmf<6D><66><EFBFBD><EFBFBD><EFBFBD><EFBFBD>R#L<>X<EFBFBD>D<EFBFBD><44><EFBFBD> <1D><><EFBFBD>Z<EFBFBD> <1C><>T<EFBFBD>UT<55><54>g<EFBFBD><67><EFBFBD>U<EFBFBD><55>:"r<>Ͻ<EFBFBD><CFBD><EFBFBD>:<3A><>H <1C>E^(<28>d [<5B><19>I<EFBFBD>I<17><><EFBFBD><EFBFBD>ݲř<DDB2><C599>3<EFBFBD><33><EFBFBD><EFBFBD><EFBFBD>N<EFBFBD><4E><01>N>
|
||||
endstream
|
||||
endobj
|
||||
94 0 obj
|
||||
4925
|
||||
endobj
|
||||
93 0 obj
|
||||
<< /Length 95 0 R /Filter /FlateDecode >>
|
||||
stream
|
||||
x<EFBFBD>k`<60><><EFBFBD><EFBFBD><EFBFBD><EFBFBD>U<<3C><> Va
|
||||
endstream
|
||||
endobj
|
||||
95 0 obj
|
||||
20
|
||||
endobj
|
||||
96 0 obj
|
||||
<<
|
||||
/Type /Font
|
||||
/Subtype /Type0
|
||||
/BaseFont /EAAAAC+mwa_cmr10
|
||||
/Encoding /Identity-H
|
||||
/ToUnicode 97 0 R
|
||||
/DescendantFonts [98 0 R]
|
||||
>>
|
||||
endobj
|
||||
98 0 obj
|
||||
<< /Type /Font
|
||||
/BaseFont /EAAAAC+mwa_cmr10
|
||||
/CIDToGIDMap /Identity
|
||||
/Subtype /CIDFontType2
|
||||
/CIDSystemInfo << /Registry (Adobe) /Ordering (UCS) /Supplement 0 >>
|
||||
/FontDescriptor 91 0 R
|
||||
/DW 0
|
||||
/W [ 0 [365 500 500 500 500 651 391 443 526 555 555 443 526 276 750 443 276 916 500 500 276 388 555 555 555 276 500 ] ]
|
||||
>>
|
||||
endobj
|
||||
97 0 obj
|
||||
<< /Length 99 0 R /Filter /FlateDecode >>
|
||||
stream
|
||||
x<EFBFBD>]<5D><>j<EFBFBD>0<10><><EFBFBD>)r<><72>X4 waʖ<>=P<><07>ɸj Q/|<7C><><EFBFBD><EFBFBD>B <1E><>?N~g<>[<5B>T<EFBFBD>~fѻ eC3<43>z<EFBFBD>,M<>b%<25><><EFBFBD><EFBFBD>><3E><><EFBFBD>^<>]<0E> "<22>ܬ<EFBFBD>LC<4C><43><EFBFBD><EFBFBD>P<EFBFBD><50>%cч{<7B>f<EFBFBD><66><EFBFBD><EFBFBD>[z`<60><>m<EFBFBD><6D>*<2A><><EFBFBD><EFBFBD><EFBFBD><EFBFBD>w<EFBFBD>Ř H<>,<2C>{<7B><>F<7F>a^<5E>@,<2C><>9<EFBFBD>ʉ<EFBFBD>y=<3D><>?<3F><>j<EFBFBD>%<25>9<EFBFBD> GE<47><11><><EFBFBD>w:\c<>*v}v<>ھ<EFBFBD>/<2F><>Hk;<3B>-<2D>.<2E>ܪ<qGq<47>rP<02>A)(e<><04>{<7B>
|
||||
P<EFBFBD><EFBFBD> <0C><>S<EFBFBD><53>Έ<EFBFBD>
|
||||
P<> ( ԂB= <09><05><<3C>-<2D>P<EFBFBD> <0C><><14>x <0C><02>_<08>
|
||||
<EFBFBD><EFBFBD><EFBFBD>_Q<EFBFBD><EFBFBD><EFBFBD>@=e<06><>"<10><>8'<0F><><EFBFBD>8<EFBFBD>%pă<70>η&<26>`k<>6l<36><04><>Z<EFBFBD>|?k<><6B>[<5B>{M<>К<EFBFBD>lY<6C><59><05><>¸
|
||||
endstream
|
||||
endobj
|
||||
99 0 obj
|
||||
347
|
||||
endobj
|
||||
72 0 obj
|
||||
<< /Type /Pages
|
||||
/Count 1
|
||||
/Kids [70 0 R ] >>
|
||||
endobj
|
||||
100 0 obj
|
||||
<<
|
||||
/Type /Catalog
|
||||
/Pages 72 0 R
|
||||
/Lang (x-unknown)
|
||||
>>
|
||||
endobj
|
||||
71 0 obj
|
||||
<<
|
||||
/Font <<
|
||||
/F1358 78 0 R
|
||||
/F1357 87 0 R
|
||||
/F1359 96 0 R
|
||||
>>
|
||||
/ProcSet [/PDF /ImageB /ImageC /Text]
|
||||
/ExtGState <<
|
||||
/GS1 2 0 R
|
||||
/GS2 3 0 R
|
||||
/GS3 4 0 R
|
||||
/GS4 5 0 R
|
||||
/GS5 6 0 R
|
||||
/GS6 7 0 R
|
||||
/GS7 8 0 R
|
||||
/GS8 9 0 R
|
||||
/GS9 10 0 R
|
||||
/GS10 11 0 R
|
||||
/GS11 12 0 R
|
||||
/GS12 13 0 R
|
||||
/GS13 14 0 R
|
||||
/GS14 15 0 R
|
||||
/GS15 16 0 R
|
||||
/GS16 17 0 R
|
||||
/GS17 18 0 R
|
||||
/GS18 19 0 R
|
||||
/GS19 20 0 R
|
||||
/GS20 21 0 R
|
||||
/GS21 22 0 R
|
||||
/GS22 23 0 R
|
||||
/GS23 24 0 R
|
||||
/GS24 25 0 R
|
||||
/GS25 26 0 R
|
||||
/GS26 27 0 R
|
||||
/GS27 28 0 R
|
||||
/GS28 29 0 R
|
||||
/GS29 30 0 R
|
||||
/GS30 31 0 R
|
||||
/GS31 32 0 R
|
||||
/GS32 33 0 R
|
||||
/GS33 34 0 R
|
||||
/GS34 35 0 R
|
||||
/GS35 36 0 R
|
||||
/GS36 37 0 R
|
||||
/GS37 38 0 R
|
||||
/GS38 39 0 R
|
||||
/GS39 40 0 R
|
||||
/GS40 41 0 R
|
||||
/GS41 42 0 R
|
||||
/GS42 43 0 R
|
||||
/GS43 44 0 R
|
||||
/GS44 45 0 R
|
||||
/GS45 46 0 R
|
||||
/GS46 47 0 R
|
||||
/GS47 48 0 R
|
||||
/GS48 49 0 R
|
||||
/GS49 50 0 R
|
||||
/GS50 51 0 R
|
||||
/GS51 52 0 R
|
||||
/GS52 53 0 R
|
||||
/GS53 54 0 R
|
||||
/GS54 55 0 R
|
||||
/GS55 56 0 R
|
||||
/GS56 57 0 R
|
||||
/GS57 58 0 R
|
||||
/GS58 59 0 R
|
||||
/GS59 60 0 R
|
||||
/GS60 61 0 R
|
||||
/GS61 62 0 R
|
||||
/GS62 63 0 R
|
||||
/GS63 64 0 R
|
||||
/GS64 65 0 R
|
||||
/GS65 66 0 R
|
||||
>>
|
||||
>>
|
||||
endobj
|
||||
xref
|
||||
Binary file not shown.
|
Before Width: | Height: | Size: 198 KiB |
Binary file not shown.
Binary file not shown.
Binary file not shown.
File diff suppressed because it is too large
Load Diff
Binary file not shown.
Binary file not shown.
Binary file not shown.
Binary file not shown.
637
mohit/paper.org
637
mohit/paper.org
@@ -1,637 +0,0 @@
|
||||
Sensor fusion , Optimal filters , $\mathcal{H}_\infty$ synthesis ,
|
||||
Vibration isolation , Precision
|
||||
|
||||
* Introduction
|
||||
The sensors used for measuring physical quantity often works well within
|
||||
a limited frequency range called as the bandwidth of the sensor. The
|
||||
signals recorded by the sensor beyond its bandwidth are often corrupt
|
||||
with noise and are not reliable. Many dynamical systems require
|
||||
measurements over a wide frequency range. Very often a variety of
|
||||
sensors are utilized to sense the same quantity. These sensors have
|
||||
different operational bandwidth and are reliable only in a particular
|
||||
frequency range. The signals from the different sensors are fused
|
||||
together in order to get the reliable measurement of the physical
|
||||
quantity over wider frequency band. The combining of signals from
|
||||
various sensor is called sensor
|
||||
fusion \cite{hua04_polyp_fir_compl_filter_contr_system}. The resulting
|
||||
sensor is referred as "super sensor" since it can have better noise
|
||||
characteristics and can operate over a wider frequency band as compared
|
||||
to the individual sensor used for
|
||||
merging \cite{shaw90_bandw_enhan_posit_measur_using_measur_accel}.
|
||||
|
||||
Sensor fusion is most commonly employed in the navigation systems to
|
||||
accurately measure the position of a vehicle. The GPS sensors, which are
|
||||
accurate in low frequency band, are merged with the high-frequency
|
||||
accelerometers. Zimmermann and
|
||||
Sulzer \cite{zimmermann92_high_bandw_orien_measur_contr} used sensor
|
||||
fusion to measure the orientation of a robot. They merged inclinometer
|
||||
and accelerometers for accurate angular measurements over large
|
||||
frequency band.
|
||||
Corke \cite{corke04_inert_visual_sensin_system_small_auton_helic} merged
|
||||
inertial measurement unit with the stereo vision system for measurement
|
||||
of attitude, height and velocity of an unmanned helicopter. Min and
|
||||
Jeung \cite{min15_compl_filter_desig_angle_estim} used accelerometer and
|
||||
gyroscope for angle estimations. Baerveldt and Klang \cite{baerveldt97}
|
||||
used an inclinometer and a gyroscope to measure the orientation of the
|
||||
autonomous helicopter. The measurement of the 3D orientation using a
|
||||
gyroscope and an accelerometer was demonstrated by Roberts et
|
||||
al. \cite{roberts03_low}. Cao et
|
||||
al. \cite{cao20_adapt_compl_filter_based_post} used sensor fusion to
|
||||
obtain the lateral and longitudinal velocities of the autonomous
|
||||
vehicle.
|
||||
|
||||
Sensor fusion is also used for enhancing the working range of the active
|
||||
isolation system. For example, the active vibration isolation system at
|
||||
the Laser Interferometer Gravitational-Wave Observatory
|
||||
(LIGO) \cite{matichard15_seism_isolat_advan_ligo} utilizes sensor
|
||||
fusion. The position sensors, seismometer and geophones are used for
|
||||
measuring the motion of the LIGO platform in different frequency
|
||||
bands \cite{hua05_low_ligo}. Tjepkema et
|
||||
al. \cite{tjepkema12_sensor_fusion_activ_vibrat_isolat_precis_equip}
|
||||
used sensor fusion to isolate precision equipment from the ground
|
||||
motion. The feedback from the accelerometer was used for active
|
||||
isolation at low frequency while force sensor was used at high
|
||||
frequency. Various configurations of sensor fusion for active vibration
|
||||
isolation systems are discussed by Collette and
|
||||
Matichard \cite{collette15_sensor_fusion_method_high_perfor}. Ma and
|
||||
Ghasemi-Nejhad \cite{ma04_frequen_weigh_adapt_contr_simul} used laser
|
||||
sensor and piezoelectric patches for simultaneous tracking and vibration
|
||||
control in smart structures. Recently, Verma et
|
||||
al. \cite{verma21_virtual_sensor_fusion_high_precis_contr} presented
|
||||
virtual sensor fusion for high precision control where the signals from
|
||||
a physical sensor are fused with a sensor simulated virtually.
|
||||
|
||||
Fusing signals from different sensors can typically be done using Kalman
|
||||
filtering \cite{odry18_kalman_filter_mobil_robot_attit_estim, ren19_integ_gnss_hub_motion_estim, faria19_sensor_fusion_rotat_motion_recon, liu18_innov_infor_fusion_method_with, abdel15_const_low_cost_gps_filter, biondi17_attit_recov_from_featur_track}
|
||||
or complementary
|
||||
filters \cite{brown72_integ_navig_system_kalman_filter}. A set of
|
||||
filters is said to be complementary if the sum of their transfer
|
||||
functions is equal to one at all frequencies. When two filters are
|
||||
complementary, usually one is a low pass filter while the other is an
|
||||
high pass filter. The complementary filters are designed in such a way
|
||||
that their magnitude is close to one in the bandwidth of the sensor they
|
||||
are combined with. This enables to measure the physical quantity over
|
||||
larger bandwidth. There are two different categories of complementary
|
||||
filters --- frequency domain complementary filters and state space
|
||||
complementary filters. Earliest application of the the frequency domain
|
||||
complementary filters was seen in Anderson and
|
||||
Fritze \cite{anderson53_instr_approac_system_steer_comput}. A simple RC
|
||||
circuit was used to physically realize the complementary filters.
|
||||
Frequency domain complementary filters were also used
|
||||
in \cite{shaw90_bandw_enhan_posit_measur_using_measur_accel, zimmermann92_high_bandw_orien_measur_contr, baerveldt97, roberts03_low}.
|
||||
State space complementary filter finds application in tracking
|
||||
orientation of the flexible links in a
|
||||
robot \cite{bachmann03_desig_marg_dof, salcudean91_global_conver_angul_veloc_obser, mahony08_nonlin_compl_filter_special_orthog_group}
|
||||
and are particularly useful for multi-input multi-output systems.
|
||||
Pascoal et al. \cite{pascoal00_navig_system_desig_using_time} presented
|
||||
complementary filters which can adapt with time for navigation system
|
||||
capable of estimating position and velocity using GPS and SONAR sensors.
|
||||
|
||||
The noise characteristics of the super sensor are governed by the norms
|
||||
of the complementary filters. Therefore, the proper design of the
|
||||
complementary filters for sensor fusion is of immense importance. The
|
||||
design of complementary filters is a complex task as they need to tuned
|
||||
as per the specification of the sensor. In many applications, analytical
|
||||
formulas of first and second order complementary filters are
|
||||
used \cite{corke04_inert_visual_sensin_system_small_auton_helic,jensen13_basic_uas}.
|
||||
These filters are easy to tune and simple to implement using an analog
|
||||
circuit \cite{moore19_capac_instr_sensor_fusion_high_bandw_nanop,cite:yong16_high_speed_vertic_posit_stage}.
|
||||
However, these low order complementary filters are not optimal, and high
|
||||
order complementary filters can lead to better
|
||||
fusion \cite{jensen13_basic_uas,shaw90_bandw_enhan_posit_measur_using_measur_accel}.
|
||||
|
||||
Several design techniques have been proposed to design higher order
|
||||
complementary filters.
|
||||
Pascoal \cite{pascoal00_navig_system_desig_using_time} used linear
|
||||
matrix inequalities (LMIs) \cite{boyd94_linear} for the design of time
|
||||
varying complementary filters. LMIs were also used by Hua et
|
||||
al. \cite{hua04_polyp_fir_compl_filter_contr_system} to design finite
|
||||
impulse response (FIR) filters for the active vibration isolation system
|
||||
at LIGO.
|
||||
Plummer \cite{plummer06_optim_compl_filter_their_applic_motion_measur}
|
||||
proposed an optimal design method using the $\mathcal{H}_{\infty}$
|
||||
synthesis and weighting functions representing the measurement noise of
|
||||
the sensors.
|
||||
|
||||
Although various methods have been presented in the literature for the
|
||||
design of complementary filters, there is a lack of general and simple
|
||||
framework that allows to shape the norm of complementary filters. Such a
|
||||
method would prove to be very useful as the noise of the "supper sensor"
|
||||
and its dynamical characteristics depend on the norm of the filters.
|
||||
This paper presents such a framework based on the $\mathcal{H}_\infty$
|
||||
norm minimization. The proposed method is quite general and can be
|
||||
easily extended to a case where more than two complementary filters
|
||||
needs to be designed. The organization of this paper is as follows.
|
||||
Section [[*Complementary filters requirements][2]] presents the design
|
||||
requirements of ideal complementary filters. It also demonstrates how
|
||||
the noise and robustness characteristics of the "super sensor" can be
|
||||
transformed into upper bounds on the norm of the complementary filters.
|
||||
The framework for the design of complementary filters is detailed in
|
||||
Section [[*Design formulation using $\mathcal{H}_\infty$ synthesis][3]].
|
||||
This is followed by the application of the design method to
|
||||
complementary filter design for the active vibration isolation at LIGO
|
||||
in Section [[*Application: Complementary Filter Design for Active
|
||||
Vibration Isolation of LIGO][4]]. Finally, concluding remarks are
|
||||
presented in Section [[*Concluding remarks][5]].
|
||||
|
||||
* Complementary filters requirements
|
||||
** Introduction :ignore:
|
||||
Complementary filters provides a framework for fusing signals from different
|
||||
sensors. As the effectiveness of the fusion depends on the proper design of the
|
||||
complementary filters, they are expected to fulfill certain requirements. These
|
||||
requirements are discussed in this section.
|
||||
|
||||
** Complementary characteristics
|
||||
Consider a case where two different sensors are used for measuring the
|
||||
same quantity, $x$ in different frequency range. The inherent dynamics
|
||||
of the sensors is represented by transfer functions $G_1(s)$ and
|
||||
$G_2(s)$. The two sensor also have uncorrelated noise characteristics
|
||||
given by $n_1$ and $n_2$. The signals from these two sensors are fused
|
||||
using complementary filters $H_1(s)$ and $H_2(s)$. The architecture of
|
||||
sensor fusion using complementary filters is shown in
|
||||
Figure [[#fig:fusion_super_sensor][1]]. The resulting sensor, termed as
|
||||
"super sensor", can have larger bandwidth and better noise
|
||||
characteristics in comparison to the individual sensor. This means that
|
||||
the super sensor provides an estimate $\hat{x}$ of $x$ which can be more
|
||||
accurate over a larger frequency band than the outputs of the individual
|
||||
sensors. Based on Figure [[#fig:fusion_super_sensor][1]], the estimate
|
||||
of the physical quantity as measured by the super sensor can be written
|
||||
as $$\label{eq:comp_filter_estimate}
|
||||
\hat{x} = \left(G_1 H_1 + G_2 H_2\right) x + H_1 n_1 + H_2 n_2$$
|
||||
|
||||
#+caption: [fig:fusion_super_sensor]Block diagram of sensor fusion with
|
||||
complementary filters
|
||||
[[file:figs/fusion_super_sensor.pdf]]
|
||||
|
||||
The complementary property of filters $H_1(s)$ and $H_2(s)$ simply
|
||||
implies that the summation of their transfer functions is equal to
|
||||
unity. That is, unity magnitude and zero phase at all frequencies
|
||||
(filters which satisfies only the magnitude condition are referred as
|
||||
"magnitude complementary filter pair"). Therefore, a pair of strict
|
||||
complementary filter needs to satisfy the following condition.
|
||||
|
||||
$$\label{eq:comp_filter}
|
||||
H_1(s) + H_2(s) = 1$$
|
||||
|
||||
** Noise characterization
|
||||
In order to compute the noise characteristics associated with the
|
||||
estimate $\hat{x}$, it is first assumed that the dynamics of the
|
||||
individual sensors are perfect: $$\label{eq:perfect_dynamics}
|
||||
G_1(s) = G_2(s) = 1$$
|
||||
|
||||
The output of the super sensor, $\hat{x}$, based on the block diagram
|
||||
shown in Figure [[#fig:fusion_super_sensor][1]] can be written as
|
||||
$$\label{eq:estimate_perfect_dyn}
|
||||
\hat{x} = x + H_1 n_1 + H_2 n_2$$
|
||||
|
||||
The complementary are operating only on the noise component of the
|
||||
individual sensor. Thus, this sensor fusion architecture permits to
|
||||
filter the noise of both sensors without introducing any distortion in
|
||||
the physical quantity to be measured. The estimation error, $\delta x$,
|
||||
of the super sensor can be written as $$\label{eq:estimate_error}
|
||||
\delta x \triangleq \hat{x} - x = H_1 n_1 + H_2 n_2$$
|
||||
|
||||
The power spectral density (PSD) of the super sensor's estimation error
|
||||
is given by $$\label{eq:noise_filtering_psd}
|
||||
\Phi_{\delta x} = \left|H_1\right|^2 \Phi_{n_1} + \left|H_2\right|^2 \Phi_{n_2}$$
|
||||
where, $\Phi_{\delta x}$ is the PSD of estimation error, $\Phi_{n_1}$
|
||||
and $\Phi_{n_2}$ are the PSDs of the noise associated with the
|
||||
individual sensor.
|
||||
It can be seen that the estimation error's PSD
|
||||
depends on the PSD of the noise in individual sensor as well as the norm
|
||||
of the complementary filters.
|
||||
Therefore, by properly shaping the norm of
|
||||
the complementary filters, it is possible to minimize the noise of the
|
||||
super sensor noise.
|
||||
|
||||
** Robustness requirements
|
||||
In the previous subsection, the inherent sensor dynamics were ignored.
|
||||
However in the real system, the sensor dynamics is not equal to unity.
|
||||
In such cases, the output of the sensor is normalized using a filter
|
||||
whose transfer function is equal to the inverse of the sensor dynamics.
|
||||
There are two major concerns in using inversion. First being the sensors
|
||||
may not have been calibrated properly and the actual sensor dynamics is
|
||||
not exactly compensated by the inverse filter. The second problem is
|
||||
that the inversion of sensor dynamics can result in an improper transfer
|
||||
function and hence may not be physically realizable. We here suppose
|
||||
that the sensor dynamics can be inverted using a proper and stable
|
||||
transfer function $\hat{G}_i(s)$. However, we suppose there exists a
|
||||
normalization error since $\hat{G}_i^{-1}(s) G_i(s) \neq 1$. This
|
||||
normalization error can be represented using frequency dependent
|
||||
multiplicative uncertainty
|
||||
(Figure [[#fig:sensor_fusion_dynamic_uncertainty][2]]). In
|
||||
Figure [[#fig:sensor_fusion_dynamic_uncertainty][2]], $\Delta_i(s)$
|
||||
satisfies $\|\Delta_i(s)\|_\infty \le 1$ and $|w_i(s)|$ is a frequency
|
||||
dependent weighting function that represents the uncertainty
|
||||
corresponding to the normalization error.
|
||||
|
||||
#+caption: [fig:sensor_fusion_dynamic_uncertainty] Representation of
|
||||
normalization error in sensor fusion using multiplicative uncertainty
|
||||
[[file:figs/sensor_fusion_dynamic_uncertainty.pdf]]
|
||||
|
||||
Based on Figure [[#fig:sensor_fusion_dynamic_uncertainty][2]], the super
|
||||
sensor dynamics can be written as
|
||||
$$\label{eq:super_sensor_dyn_uncertainty}
|
||||
\frac{\hat{x}}{x} = 1 + w_1(s) H_1(s) \Delta_1(s) + w_2(s) H_2(s) \Delta_2(s)$$
|
||||
|
||||
The dynamics of the super sensor now depends on the weighting functions
|
||||
($w_1(s),w_2(s)$) and the complementary filters ($H_1(s),H_2(s)$).
|
||||
|
||||
The robust stability of the fusion can be studied graphically (refer
|
||||
Figure [[#fig:uncertainty_set_super_sensor][3]]). The frequency response of the fusion output is plotted in a complex plane. The unity transfer function leads to a point $(1,0)$ located on the real axis. The uncertainty associated with first sensor at a particular frequency is represented by a circle with the center at (1,0) and radius $|w_1H_1|$. The uncertainty associated with the second is also represented using a circle centered at any point on the circle representing uncertainty associated with the first sensor and radius equal to $|w_2H_2|$. Therefore, the overall uncertainty of the fusion is represented with a circle centered at (1,0) and radius equal to $|w_1H_1|+|w_2H_2|$.
|
||||
Mathematically, the maximum phase difference at frequency
|
||||
$\omega$ that can result from fusion is given by
|
||||
$$\label{eq:max_phase_uncertainty}
|
||||
\Delta\phi(\omega) = \arcsin\left( |w_1(j\omega) H_1(j\omega)| + |w_2(j\omega) H_2(j\omega)| \right)$$
|
||||
|
||||
#+caption: [fig:uncertainty_set_super_sensor]Robustness analysis of
|
||||
sensor fusion in the complex plane. The uncertainty associated with the
|
||||
super sensor dynamics are represented with a solid circle while those
|
||||
associated with individual sensors are represented with dashed circles.
|
||||
[[file:figs/uncertainty_set_super_sensor.pdf]]
|
||||
|
||||
A constraint on the maximum phase difference at a given frequency can be
|
||||
incorporated in the design problem using the following equation:
|
||||
$$\label{eq:max_uncertainty_super_sensor}
|
||||
\max_\omega \big( \left|w_1 H_1\right| + \left|w_2 H_2\right|\big) < \sin\left( \Delta \phi_\text{max} \right)$$
|
||||
where $\Delta \phi_\text{max}$ is the maximum allowable phase
|
||||
difference. It can also be inferred from the above equation that the
|
||||
magnitude of the complementary filter ($|H_i|$) should be tuned to a
|
||||
smaller value at the frequencies where the magnitude of weighing
|
||||
transfer functions ($|w_i|$) representing sensor uncertainty is large.
|
||||
|
||||
* Design formulation using $\mathcal{H}_\infty$ synthesis
|
||||
** Introduction :ignore:
|
||||
In this section, the shaping of complementary filters is expressed as an optimal $\mathcal{H}_{\infty}$ synthesis problem.
|
||||
The synthesis goal is to shape the frequency response of the filters such that they satisfy the design requirements presented in Section [[*Complementary filters
|
||||
requirements][2]].
|
||||
|
||||
** Synthesis problem formulation
|
||||
The first step is to formulate the filter design problem as a
|
||||
generalized plant-controller structure \cite{boyd91_linear}. The
|
||||
generalized plant and controller structure for complementary filters
|
||||
design is shown in Figure [[#fig:h_infinity_robust_fusion][4]]. In the
|
||||
figure, $P(s)$ is the generalized plant, $u$ is the "control input", $v$
|
||||
is the "measured output" and $H_2(s)$ is the controller (filter) to be
|
||||
designed. The regulated outputs of the generalized plant, $z_1$ and
|
||||
$z_2$, are given by $$\begin{split}
|
||||
z_1 &= W_1(s)(1-H_2(s)) w = W_1(s) H_1(s) w \text{ by defining } H_{1}(s) \triangleq 1 - H_{2}(s)\\
|
||||
z_2 &= W_2(s) H_2(s) w
|
||||
\end{split}$$ where $w$ is the "exogenous input" to the plant,
|
||||
$W_1(s), W_2(s)$ are the weighting functions for shaping the
|
||||
complementary filters.
|
||||
|
||||
The dynamics of the generalized plant can be written as
|
||||
$$\label{eq:generalized_plant}
|
||||
\begin{bmatrix} z_1 \\ z_2 \\ v \end{bmatrix} = P(s) \begin{bmatrix} w\\u \end{bmatrix}; \quad P(s) = \begin{bmatrix}W_1(s) & -W_1(s) \\ 0 & W_2(s) \\ 1 & 0 \end{bmatrix}$$
|
||||
|
||||
The weighting functions are chosen based on the specifications and
|
||||
requirements set for the complementary filters (discussed in
|
||||
Section [[*Design of weighting functions][3.2]]). The objective of the
|
||||
optimization is to design a filter $H_2(s)$ such that the following
|
||||
conditions are satisfied $$\label{eq:comp_filter_problem_form}
|
||||
\begin{split}
|
||||
\left| \frac{z_{1}}{w} \right| &= |1-H_2(s)| \le \frac{1}{|W_1(s)|} \\
|
||||
\left| \frac{z_{1}}{w} \right| &= |H_2(s)| \le \frac{1}{|W_2(s)|}
|
||||
\end{split}, \quad \forall \omega \in \mathbb{R}^{+}$$
|
||||
|
||||
#+caption: [fig:h_infinity_robust_fusion]Generalized plant controller
|
||||
structure for the design of complementary filters
|
||||
[[file:figs/h_infinity_robust_fusion.pdf]]
|
||||
|
||||
Based on Figure [[#fig:h_infinity_robust_fusion][4]], the
|
||||
$\mathcal{H}_{\infty}$ synthesis problem for the complementary filters
|
||||
can be stated as
|
||||
|
||||
#+begin_quote
|
||||
Find a stable transfer function, $H_2(s)$, which takes measured
|
||||
output, $v$, as input and generates a control input, $u$, such that
|
||||
the $\mathcal{H}_\infty$ norm of the generalized plant from exogenous
|
||||
input, $w$, to the regulated output, ${[z_1,z_2]}^T$ is less than
|
||||
unity.
|
||||
#+end_quote
|
||||
|
||||
Mathematically, the synthesis objective can be written as
|
||||
$$\begin{split}
|
||||
&\left\|\begin{matrix} \left[1 - H_2(s)\right] W_1(s) \\ H_2(s) W_2(s) \end{matrix}\right\|_\infty \le 1 \\
|
||||
\Longleftrightarrow & \left\|\begin{matrix} H_1(s) W_1(s) \\ H_2(s) W_2(s) \end{matrix}\right\|_\infty \le 1; \quad H_1(s) \triangleq 1 - H_2(s)
|
||||
\end{split}
|
||||
\label{eq:hinf_syn_obj}$$ The above optimization problem can be
|
||||
efficiently solved in Matlab \cite{MATLAB2009} using Riccati formulae,
|
||||
linear matrix inequality based method or maximum entropy method.
|
||||
|
||||
** Design of weighting functions
|
||||
The choice of weighting function governs the shape of the designed
|
||||
complementary filters. Therefore, it is very important that the design
|
||||
specifications are appropriately transformed into the weighting
|
||||
functions. The choice of weighting functions is also constrained by the
|
||||
following factors
|
||||
|
||||
1. Only proper and stable transfer functions can be used as weighting
|
||||
functions
|
||||
|
||||
2. As the order of the designed filter is equal to the sum of the orders
|
||||
of the weighting functions, the order of the weighting function needs
|
||||
to be reasonably small to ensure the physical implementation of the
|
||||
designed complementary filters. This also reduces the computational
|
||||
cost of the optimization problem.
|
||||
|
||||
3. The complementary property of the filter imposes a fundamental
|
||||
limitations on the weighting functions. The imposes a restriction
|
||||
that the magnitude of the filters $H_1(s)$ and $H_2(s)$ cannot be
|
||||
made small simultaneously at the same frequency.
|
||||
|
||||
The specifications of the complementary filters are typically expressed
|
||||
using the following parameters --- low frequency gain, high frequency
|
||||
gain, slope (order of the filter) and the crossover frequency. We
|
||||
propose a weighting function that allows to translate the above
|
||||
requirements by setting simple parameters: $$\label{eq:weight_formula}
|
||||
W(s) = \displaystyle\left( \frac{
|
||||
\hfill{} \displaystyle\frac{1}{\omega_0} \sqrt{\frac{1 - \left(\displaystyle\frac{G_0}{G_c}\right)^{\displaystyle\frac{2}{n}}}{1 - {\left(\displaystyle\frac{G_c}{G_\infty}\right)}^{\displaystyle\frac{2}{n}}}} s + \left(\frac{G_0}{G_c}\right)^{\displaystyle\frac{1}{n}}
|
||||
}{
|
||||
\left(\displaystyle\frac{1}{G_\infty}\right)^{\displaystyle\frac{1}{n}}\displaystyle \frac{1}{\omega_0} \sqrt{\displaystyle\frac{1 - \left(\displaystyle\frac{G_0}{G_c}\right)^{\displaystyle\frac{2}{n}}}{1 - \left(\displaystyle\frac{G_c}{G_\infty}\right)^{\displaystyle\frac{2}{n}}}} s + \left(\displaystyle\frac{1}{G_c}\right)^{\displaystyle\frac{1}{n}}
|
||||
}\right)^n$$ where, $G_0 = \lim_{\omega \to 0} |W(j\omega)|$ is the
|
||||
low frequency gain, $G_\infty = \lim_{\omega \to \infty} |W(j\omega)|$
|
||||
is the high frequency gain, $\omega_c$ is the crossover frequency,
|
||||
$G_c = |W(j\omega_c)|$ is the absolute gain at the crossover frequency
|
||||
and $n$ is the order of the filter. As an illustration, the magnitude of
|
||||
the frequency response of the weighting function with the parameters
|
||||
$G_0 = 0.001$, $G_\infty = 10$, $\omega_c = \SI{10}{Hz}$, $G_c = 2$,
|
||||
$n = 3$ and having high pass characteristics is shown in
|
||||
Figure [[#fig:weight_formula][5]].
|
||||
|
||||
#+caption: [fig:weight_formula]Magnitude of the frequency response of
|
||||
the weighting function obtained using
|
||||
equation [[#eq:weight_formula][[eq:weight_formula]]] with the parameters
|
||||
$G_0 = 0.001$, $G_\infty = 10$, $\omega_c = \SI{10}{Hz}$, $G_c = 2$,
|
||||
$n = 3$
|
||||
[[file:figs/weight_formula.pdf]]
|
||||
|
||||
** Verification
|
||||
The proposed methodology for the design of complementary filters is now
|
||||
applied on the following example.
|
||||
|
||||
*Example.* Design complementary filters with the merging frequency
|
||||
around 10 Hz. The low pass filter should have DC gain of 0.001 and slope
|
||||
-2 above the merging frequency. The high pass filter should have a slope
|
||||
of 3 below the merging frequency and 0.001 high frequency gain.
|
||||
|
||||
The first step is to design the weighting functions that translate the
|
||||
above requirements. To do so,
|
||||
equation [[#eq:weight_formula][[eq:weight_formula]]] is used. The
|
||||
parameters corresponding to $W_1(s)$ and $W_2(s)$ are listed in
|
||||
Table [[#tab:weights_params][1]]. The obtained transfer functions of the
|
||||
weighting functions are $$\begin{split}
|
||||
W_1(s) &= \dfrac{1000 (s+34.55)^2}{(s+3455)^2}\\
|
||||
W_2(s) &= \dfrac{0.1 (s+87.43)^3}{(s+4.058)^3}
|
||||
\end{split}$$
|
||||
|
||||
Using these weighting functions, the generalized plant is evaluated
|
||||
using equation [[#eq:generalized_plant][[eq:generalized_plant]]]. The
|
||||
optimal complementary filters are obtained by solving the optimization
|
||||
problem given by equation [[#eq:hinf_syn_obj][[eq:hinf_syn_obj]]]. The
|
||||
complementary filters obtained after optimization are $$\begin{split}
|
||||
H_1(s) &= \frac{10^{-8} (s+6.6\times 10^9) (s+3450)^2 (s^2 + 49s + 895)}{(s+6.6e^4) (s^2 + 106 s + 3\times 10^3) (s^2 + 72s + 3580)}\\
|
||||
H_2(s) &= \frac{(s+6.6\times 10^4) (s+160) (s+4)^3}{(s+6.6\times 10^4) (s^2 + 106 s + 3\times 10^3) (s^2 + 72s + 3580)}
|
||||
\end{split}$$
|
||||
|
||||
The obtained complementary filters are of order 5 which corresponds to
|
||||
the sum of the orders of the weighting functions used. The frequency
|
||||
responses of the designed complementary filters are shown in
|
||||
Figure [[#fig:hinf_filters_results][6]]. It can be seen that the
|
||||
designed filters fulfills all the design specifications and hence
|
||||
demonstrates the effectiveness of the designed methodology (more complex
|
||||
real life example is taken up in Section [[*Application: Complementary
|
||||
Filter Design for Active Vibration Isolation of LIGO][4]]).
|
||||
|
||||
<<tab:weights_params>>
|
||||
| Parameter | $W_1(s)$ | $W_2(s)$ |
|
||||
|------------------------+----------+----------|
|
||||
| $n$ | $2$ | $3$ |
|
||||
| $G_c$ | $0.5$ | $0.5$ |
|
||||
| $G_\infty$ | $1000$ | $0.1$ |
|
||||
| $G_0$ | $0.1$ | $1000$ |
|
||||
| $\omega_c$ [$\si{Hz}$] | $11$ | $10$ |
|
||||
#+caption: [tab:weights_params]Parameters used for $W_1(s)$ and $W_2(s)$
|
||||
|
||||
#+caption: [fig:hinf_filters_results]Bode plot of the weighting
|
||||
functions and designed complementary filters
|
||||
[[file:figs/hinf_filters_results.pdf]]
|
||||
|
||||
** TODO Extension to set of three complementary filters
|
||||
In certain applications, more than two sensors are used to measure the
|
||||
same quantity and can be merged together to form a "super sensor". In
|
||||
such case, a set of three (or more) complementary filters is required.
|
||||
It is here shown that the proposed method can be generalized for the
|
||||
design of a set of arbitrary number of complementary filters. The
|
||||
control objective is now to design of a set of $n$ complementary filters
|
||||
($H_i(s), i=1,\cdots,n$) which satisfy the following conditions
|
||||
$$\label{eq:hinf_problem_gen}
|
||||
\begin{split}
|
||||
&\sum_{i=0}^n H_i(s) = 1 \\
|
||||
&\left| H_i(s) \right| < \frac{1}{\left| W_i(s) \right|}
|
||||
\end{split}$$ Here, we extend the method to a case of three
|
||||
complementary filters. The generalized plant controller setup for this
|
||||
case is shown in Figure [[#fig:comp_filter_three_hinf][7]]. The
|
||||
synthesis objective is to design filters $H_2(s)$ and $H_3(s)$ such that
|
||||
the $\mathcal{H}_\infty$ norm from exogenous input $w$ to regulated
|
||||
output vector $[z_1,z_2,z_3]^T$ is less than unity. That is,
|
||||
$$\label{eq:hinf_syn_obj_three}
|
||||
\begin{split}
|
||||
&\left\| \begin{matrix} \left[1 - H_2(s) - H_3(s)\right] W_1(s) \\ H_2(s) W_2(s) \\ H_3(s) W_3(s) \end{matrix} \right\|_\infty \le 1\\
|
||||
\equiv &\left\| \begin{matrix} H_1(s) W_1(s) \\ H_2(s) W_2(s) \\ H_3(s) W_3(s) \end{matrix} \right\|_\infty \le 1; \quad H_1(s) \triangleq 1 - H_2(s) - H_3(s)
|
||||
\end{split}$$
|
||||
|
||||
#+caption: [fig:comp_filter_three_hinf]Generalized plant controller
|
||||
setup for designing a set of three complementary filters using
|
||||
$\mathcal{H}_\infty$ synthesis
|
||||
[[file:figs/comp_filter_three_hinf.pdf]]
|
||||
|
||||
To validate this synthesis method, let's take an example where 3 sensors
|
||||
are merged together. The three sensors are working in the following
|
||||
three frequency range --- below 1 Hz, between 1Hz to 10Hz and above
|
||||
10Hz. The weighting functions used for the synthesis are $$\begin{split}
|
||||
W_1(s) &= \dfrac{1000 (s+3.141)^2}{(s+314.1)^2}\\
|
||||
W_2(s) &= \dfrac{2200 (s+62.83)^2 (s+6.283)^2}{(s+6283)^2 (s+0.06283)^2}\\
|
||||
W_3(s) &= \dfrac{0.1 (s+87.43)^3}{(s+4.058)^3}
|
||||
\end{split}$$ The complementary filters are obtained by solving the
|
||||
optimization problem given by
|
||||
equation [[#eq:hinf_syn_obj_three][[eq:hinf_syn_obj_three]]]. The
|
||||
frequency response of the designed filters and the weighting functions
|
||||
are shown in Figure [[#fig:three_complementary_filters_results][8]].
|
||||
|
||||
#+caption: [fig:three_complementary_filters_results]Bode plot of the
|
||||
weighting functions and designed set of three complementary filters
|
||||
[[file:figs/three_complementary_filters_results.pdf]]
|
||||
|
||||
* Application: Complementary Filter Design for Active Vibration
|
||||
** Introduction :ignore:
|
||||
Isolation of LIGO Gravitational waves can help in detection various astrophysical events occurring in our universe. This can also pave a path to validate theories built around the existence of gravitational waves. However, the detection of these waves is an arduous task owing to the extraordinary small strain experienced by the earth due to gravitational waves. Various methods have been proposed for their detection, out of which laser interferometers are the most popular ones. Laser interferometers offers large projection range and high displacement sensitivity. Among the existing detector, Laser interferometer gravitation-wave observatory (LIGO) is the most sensitive operational detector. LIGO consists of two longs arms, referred as beam tubes, that are placed orthogonal to each other. The arms of the LIGO accommodates a Michleson interferometer with a cavity (Fabry-Perot). The mirrors at the extremity of the cavity serve
|
||||
as inertial test masses which responds to the strain induced due to the
|
||||
gravitational waves. The optics of the LIGO are suspended like a
|
||||
pendulum. The schematics of the LIGO are shown in
|
||||
Figure [[#ligo_schematics][9]].
|
||||
|
||||
#+caption: [ligo_schematics]Schematics of LIGO
|
||||
[[file:figs/ligo.png]]
|
||||
|
||||
The isolation of the terrestrial interferometers is necessary in order
|
||||
to isolate the motion of the suspended inertial masses from the seismic
|
||||
ground motion. The vibration isolation system attenuates the effect of
|
||||
disturbances on the motion of the suspended masses in the gravitational
|
||||
wave band. The other sources that can affect the sensitivity of the LIGO
|
||||
are thermal noise, human activities, tidal motion etc. The suspensions
|
||||
of the LIGO needs to serve two purpose --- isolation and alignment. The
|
||||
alignment is also important in order to ensure that the interferometer
|
||||
beam is targeted at the center of the suspended mass. The current
|
||||
vibration isolation system for LIGO has seven different stages. In the
|
||||
first stage, hydraulically actuated external pre-isolators are used for
|
||||
attenuating large ground motions. Two stages of active electromagnetic
|
||||
isolation system are placed next to pre-isolators. This is followed by
|
||||
four stages of pendulum based passive isolation system.
|
||||
|
||||
In the active isolation stage of LIGO, different sensors are used to
|
||||
sense the same physical signal in different frequency range. For
|
||||
example, seismometers are used to sense the position of the platform in
|
||||
the frequency band 0.5--10 Hz while geophones are employed above 10 Hz.
|
||||
The signals recorded from different sensor are fused using complementary
|
||||
filters \cite{hua05_low_ligo,hua04_polyp_fir_compl_filter_contr_system}.
|
||||
The stringent requirements on these filters complicate their design.
|
||||
Hua \cite{hua05_low_ligo} proposed complementary FIR filters which were
|
||||
synthesized using convex optimization. The designed FIR filters were
|
||||
found to be compliant with the design specifications. However, the order
|
||||
of the designed filter was very high, which limits its application to a
|
||||
practical system. In this section, we demonstrate the design of
|
||||
complementary filters with the same specification using the proposed
|
||||
method based on $\mathcal{H}_\infty$- synthesis.
|
||||
|
||||
** Design specifications
|
||||
The design specification of the complementary filters (as listed out
|
||||
in \cite{hua05_low_ligo}) are as follows:
|
||||
|
||||
1. In the frequency range $0$-$\SI{0.008}{Hz}$: the high pass filter's
|
||||
magnitude should be less than $8 \times 10^{-4}$.
|
||||
|
||||
2. For frequency range $\SI{0.008}{Hz}$-$\SI{0.04}{Hz}$: slope of the
|
||||
high pass filter is equal to three.
|
||||
|
||||
3. Between $\SI{0.04}{Hz}$-$\SI{0.1}{Hz}$ frequency range: the high pass
|
||||
filter's magnitude should be less than $3$.
|
||||
|
||||
4. For frequencies above $\SI{0.1}{Hz}$: the low pass filter's magnitude
|
||||
should be less than $0.045$.
|
||||
|
||||
The specification of the complementary filters are shown graphically by
|
||||
dashed black lines in Figure [[#fig:ligo_weights][10]].
|
||||
|
||||
** Weighting Functions Design
|
||||
As the synthesis objective of the complementary filters is described by
|
||||
Eq. [[#eq:hinf_problem_gen][[eq:hinf_problem_gen]]], it is clear that
|
||||
the weighting functions should be chosen such that their inverse
|
||||
magnitude represent the maximum allowed norm of the complementary
|
||||
filters. This can be done manually using by combining poles and zeros or
|
||||
using useful formulas such as
|
||||
Eq. [[#eq:weight_formula][[eq:weight_formula]]]. It is important to note
|
||||
that the order of the filters should be kept reasonable small in order
|
||||
to keep the computational cost of the optimization reasonable. This will
|
||||
also ensure that the designed filters are realizable in the physical
|
||||
world. The transfer function representing weights should also be stable
|
||||
and minimum phase.
|
||||
|
||||
The weighting function corresponding to the low pass filter, $w_L(s)$,
|
||||
is here taken as Type I Chebyshev filter. The order of the weighting
|
||||
function for low pass filter is set as 20.
|
||||
|
||||
The weighting function for the high pass filter, $w_H(s)$, is designed
|
||||
in such a way that its magnitude response is as close as possible to the
|
||||
design specifications. This was achieved using a combination of high-,
|
||||
low- and band-pass filters in the particular frequency band. The overall
|
||||
order of the weighting function for high pass filter is 7.
|
||||
|
||||
The magnitude responses of the inverse of the designed weighting
|
||||
functions and their comparison with the specifications are shown in
|
||||
Figure [[#fig:ligo_weights][10]]. It can be seen that the inverse of the
|
||||
designed weights, shown in solid blue line for high pass filter and
|
||||
solid red line for low pass filter, are close to the specifications
|
||||
shown in black dotted line.
|
||||
|
||||
#+caption: [fig:ligo_weights]Specifications and weighting functions
|
||||
magnitudes
|
||||
[[file:figs/ligo_weights.pdf]]
|
||||
|
||||
** $\mathcal{H}_\infty$ synthesis of complementary filters
|
||||
The complementary filters are designed using $\mathcal{H}_\infty$
|
||||
synthesis based on the architecture shown in
|
||||
Figure [[#fig:h_infinity_robust_fusion][4]]. The generalized plant is
|
||||
obtained by substituting the transfer functions of the designed
|
||||
weighting functions in
|
||||
equation [[#eq:generalized_plant][[eq:generalized_plant]]], i.e.,
|
||||
$W_1(s)=w_H(s)$ and $W_2(s)=w_L(s)$. The objective of the
|
||||
$\mathcal{H}_\infty$ synthesis is given by
|
||||
equation [[#eq:hinf_syn_obj][[eq:hinf_syn_obj]]]. The optimization
|
||||
problem in solved in Matlab \cite{MATLAB2009} using Ricatti method. The
|
||||
frequency responses of the designed optimal complementary filters are
|
||||
shown in Figure [[#fig:comp_fir_ligo_hinf][11]]. The order of the
|
||||
filters obtained using $\mathcal{H}_\infty$ optimization is $27$.
|
||||
|
||||
Now, we compare the designed complementary filters with the FIR filters
|
||||
designed by Hua \cite{hua05_low_ligo}. The transfer function of the FIR
|
||||
filter, $G(\omega)$, is the Fourier transform of its coefficients,
|
||||
$g(n)$. That is, $$G(\omega) = \sum_n g(n)e^{-j2n\pi\omega}$$ The FIR
|
||||
filter synthesis was formulated as a convex optimization problem. The
|
||||
objective of the optimization problem was to find the filter's
|
||||
coefficients such that their norm are below the specified upper bounds.
|
||||
The optimization problem was solved using
|
||||
SeDuMi \cite{sturm99_using_sedum} and the obtained order for the FIR
|
||||
filters is $512$. The bode plot of the FIR filters are shown with dotted
|
||||
lines in Figure [[#fig:comp_fir_ligo_hinf][11]]. It can be seen that
|
||||
frequency responses of the designed complementary filters matches quite
|
||||
well with those of the FIR filters. The designed complementary filters
|
||||
are of much lower order and can be implemented with less computational
|
||||
cost can the FIR filters. The proposed methodology for the design of
|
||||
complementary filters can be effectively employed to obtain physically
|
||||
realizable filters.
|
||||
|
||||
#+caption: [fig:comp_fir_ligo_hinf]Bode plot of the filters designed
|
||||
using $\mathcal{H}_\infty$ synthesis and FIR
|
||||
filters \cite{hua05_low_ligo}
|
||||
[[file:figs/comp_fir_ligo_hinf.pdf]]
|
||||
|
||||
* Concluding remarks
|
||||
The measurements from the sensors are reliable only within its
|
||||
bandwidth. The signals from different sensors are usually fused in order
|
||||
to measure a physical quantity over larger bandwidth. The sensor
|
||||
obtained after fusion is called as super sensor as it has superior noise
|
||||
characteristics and wider bandwidth. Complementary filters are used for
|
||||
the combining the signals from different sensors. A new framework based
|
||||
on $\mathcal{H}_\infty$ synthesis has been presented in this paper to
|
||||
aid the design of complementary filters. The method presented allows to
|
||||
shape the complementary filters based on the design specifications. The
|
||||
task of filter design is posed as an $\mathcal{H}_{\infty}$ synthesis
|
||||
problem. The design specifications of the systems are transformed in the
|
||||
form of weighting functions. These weighting functions are used in the
|
||||
optimization problem to constraint the filter response in a frequency
|
||||
band. The method has also been demonstrated for designing a set of three
|
||||
complementary filters. The design frame is general, simple to implement
|
||||
and can easily be extended to difference scenarios of sensor fusion. The
|
||||
effectiveness of the method is demonstrated for a real life application
|
||||
where complementary filters are designed for active vibration isolation
|
||||
of Laser Interferometer Gravitational-Wave Observatory (LIGO). The
|
||||
filters designed with the proposed method have been with compared with
|
||||
the finite impulse response (FIR) filters. It was found that the filters
|
||||
designed using $\mathcal{H}_\infty$ have lower order compared to FIR
|
||||
filters. The designed filters are physically realizable and have lesser
|
||||
computational cost compared to FIR filters. The proposed method can be
|
||||
effectively used to shape complementary filters based on design
|
||||
specifications. The method can be further be extended for the design of
|
||||
robust complementary filters with desired noise characteristics
|
||||
considering uncertainties in the sensor dynamics. This is the focus of
|
||||
our future research.
|
||||
|
||||
* Acknowledgment
|
||||
:PROPERTIES:
|
||||
:CLASS: unnumbered
|
||||
:END:
|
||||
The authors would like to acknowledge the help received from the French
|
||||
Community of Belgium for funding the FRIA Grant of Thomas Dehaeze (Grant
|
||||
No. FC 31597).
|
||||
BIN
mohit/paper.pdf
BIN
mohit/paper.pdf
Binary file not shown.
560
mohit/paper.tex
560
mohit/paper.tex
@@ -1,560 +0,0 @@
|
||||
\documentclass[sort&compress, review,12pt,times]{elsarticle}
|
||||
|
||||
\usepackage[utf8]{inputenc}
|
||||
\usepackage[T1]{fontenc}
|
||||
\usepackage{graphicx}
|
||||
\usepackage{grffile}
|
||||
\usepackage{rotating}
|
||||
\usepackage[normalem]{ulem}
|
||||
\usepackage{capt-of}
|
||||
\usepackage{hyperref}
|
||||
\usepackage{bm}
|
||||
\usepackage{array}
|
||||
\usepackage{amsmath,amssymb,amsfonts}
|
||||
\usepackage{algorithmic}
|
||||
\usepackage{textcomp}
|
||||
\usepackage{xcolor}
|
||||
\usepackage{cases}
|
||||
\usepackage{tabularx,siunitx,booktabs}
|
||||
\usepackage{algorithmic}
|
||||
\newcounter{example}[section]
|
||||
\newenvironment{example}[1][]{\refstepcounter{example}\par\medskip
|
||||
\noindent \textbf{Example. #1} \rmfamily}{\medskip}
|
||||
\usepackage[top=3cm, bottom=3cm, left=3cm, right=3cm]{geometry}
|
||||
|
||||
\journal{Mechanical Systems and Signal Processing}
|
||||
|
||||
\author[a1,a2]{Thomas Dehaeze\corref{cor1}}
|
||||
\author[a3,a4]{Mohit Verma}
|
||||
\author[a2,a4]{Christophe Collette}
|
||||
\cortext[cor1]{Corresponding author. Email Address: dehaeze.thomas@gmail.com}
|
||||
|
||||
\address[a1]{European Synchrotron Radiation Facility, Grenoble, France}
|
||||
\address[a2]{University of Li\`{e}ge, Department of Aerospace and Mechanical Engineering, 4000 Li\`{e}ge, Belgium.}
|
||||
\address[a3]{CSIR --- Structural Engineering Research Centre, Taramani, Chennai --- 600113, India.}
|
||||
\address[a4]{Universit\'{e} Libre de Bruxelles, Precision Mechatronics Laboratory, BEAMS Department, 1050 Brussels, Belgium.}
|
||||
|
||||
\date{}
|
||||
\title{A new method of designing complementary filters for sensor fusion using \(\mathcal{H}_\infty\) synthesis}
|
||||
|
||||
|
||||
\begin{document}
|
||||
|
||||
\begin{frontmatter}
|
||||
\begin{abstract}
|
||||
Sensor have limited bandwidth and are accurate only in a certain frequency band.
|
||||
In many applications, the signals of different sensor are fused together in order to either enhance the stability or improve the operational bandwidth of the system.
|
||||
The sensor signals can be fused using complementary filters.
|
||||
The tuning of complementary filters is a complex task and is the subject of this paper.
|
||||
The filters needs to meet design specifications while satisfying the complementary property.
|
||||
This paper presents a framework to shape the norm of complementary filters using the $\mathcal{H}_\infty$ norm minimization.
|
||||
The design specifications are imposed as constraints in the optimization problem by appropriate selection of weighting functions.
|
||||
The proposed method is quite general and easily extendable to cases where more than two sensors are fused.
|
||||
Finally, the proposed method is applied to the design of complementary filter design for active vibration isolation of the Laser Interferometer Gravitation-wave Observatory (LIGO).
|
||||
\end{abstract}
|
||||
|
||||
\begin{keyword}
|
||||
Sensor fusion \sep{} Optimal filters \sep{} $\mathcal{H}_\infty$ synthesis \sep{} Vibration isolation \sep{} Precision
|
||||
\end{keyword}
|
||||
\end{frontmatter}
|
||||
|
||||
\section{Introduction}\label{sec:introduction}
|
||||
|
||||
The sensors used for measuring physical quantity often works well within a limited frequency range called as the bandwidth of the sensor.
|
||||
The signals recorded by the sensor beyond its bandwidth are often corrupt with noise and are not reliable.
|
||||
Many dynamical systems require measurements over a wide frequency range.
|
||||
Very often a variety of sensors are utilized to sense the same quantity.
|
||||
These sensors have different operational bandwidth and are reliable only in a particular frequency range.
|
||||
The signals from the different sensors are fused together in order to get the reliable measurement of the physical quantity over wider frequency band.
|
||||
The combining of signals from various sensor is called sensor fusion~\cite{hua04_polyp_fir_compl_filter_contr_system}.
|
||||
The resulting sensor is referred as ``super sensor'' since it can have better noise characteristics and can operate over a wider frequency band as compared to the individual sensor used for merging~\cite{shaw90_bandw_enhan_posit_measur_using_measur_accel}.
|
||||
|
||||
Sensor fusion is most commonly employed in the navigation systems to accurately measure the position of a vehicle.
|
||||
The GPS sensors, which are accurate in low frequency band, are merged with the high-frequency accelerometers.
|
||||
Zimmermann and Sulzer~\cite{zimmermann92_high_bandw_orien_measur_contr} used sensor fusion to measure the orientation of a robot.
|
||||
They merged inclinometer and accelerometers for accurate angular measurements over large frequency band.
|
||||
Corke~\cite{corke04_inert_visual_sensin_system_small_auton_helic} merged inertial measurement unit with the stereo vision system for measurement of attitude, height and velocity of an unmanned helicopter.
|
||||
Min and Jeung~\cite{min15_compl_filter_desig_angle_estim} used accelerometer and gyroscope for angle estimations.
|
||||
Baerveldt and Klang~\cite{baerveldt97} used an inclinometer and a gyroscope to measure the orientation of the autonomous helicopter.
|
||||
The measurement of the 3D orientation using a gyroscope and an accelerometer was demonstrated by Roberts et al.~\cite{roberts03_low}.
|
||||
Cao et al.~\cite{cao20_adapt_compl_filter_based_post} used sensor fusion to obtain the lateral and longitudinal velocities of the autonomous vehicle.
|
||||
|
||||
Sensor fusion is also used for enhancing the working range of the active isolation system.
|
||||
For example, the active vibration isolation system at the Laser Interferometer Gravitational-Wave Observatory (LIGO)~\cite{matichard15_seism_isolat_advan_ligo} utilizes sensor fusion.
|
||||
The position sensors, seismometer and geophones are used for measuring the motion of the LIGO platform in different frequency bands~\cite{hua05_low_ligo}.
|
||||
Tjepkema et al.~\cite{tjepkema12_sensor_fusion_activ_vibrat_isolat_precis_equip} used sensor fusion to isolate precision equipment from the ground motion.
|
||||
The feedback from the accelerometer was used for active isolation at low frequency while force sensor was used at high frequency.
|
||||
Various configurations of sensor fusion for active vibration isolation systems are discussed by Collette and Matichard~\cite{collette15_sensor_fusion_method_high_perfor}.
|
||||
Ma and Ghasemi-Nejhad~\cite{ma04_frequen_weigh_adapt_contr_simul} used laser sensor and piezoelectric patches for simultaneous tracking and vibration control in smart structures.
|
||||
Recently, Verma et al.~\cite{verma21_virtual_sensor_fusion_high_precis_contr} presented virtual sensor fusion for high precision control where the signals from a physical sensor are fused with a sensor simulated virtually.
|
||||
|
||||
Fusing signals from different sensors can typically be done using Kalman filtering~\cite{odry18_kalman_filter_mobil_robot_attit_estim, ren19_integ_gnss_hub_motion_estim, faria19_sensor_fusion_rotat_motion_recon, liu18_innov_infor_fusion_method_with, abdel15_const_low_cost_gps_filter, biondi17_attit_recov_from_featur_track} or complementary filters~\cite{brown72_integ_navig_system_kalman_filter}.
|
||||
A set of filters is said to be complementary if the sum of their transfer functions is equal to one at all frequencies.
|
||||
When two filters are complementary, usually one is a low pass filter while the other is an high pass filter.
|
||||
The complementary filters are designed in such a way that their magnitude is close to one in the bandwidth of the sensor they are combined with.
|
||||
This enables to measure the physical quantity over larger bandwidth.
|
||||
There are two different categories of complementary filters --- frequency domain complementary filters and state space complementary filters.
|
||||
Earliest application of the the frequency domain complementary filters was seen in Anderson and Fritze~\cite{anderson53_instr_approac_system_steer_comput}.
|
||||
A simple RC circuit was used to physically realize the complementary filters.
|
||||
Frequency domain complementary filters were also used in~\cite{shaw90_bandw_enhan_posit_measur_using_measur_accel, zimmermann92_high_bandw_orien_measur_contr, baerveldt97, roberts03_low}.
|
||||
State space complementary filter finds application in tracking orientation of the flexible links in a robot~\cite{bachmann03_desig_marg_dof, salcudean91_global_conver_angul_veloc_obser, mahony08_nonlin_compl_filter_special_orthog_group} and are particularly useful for multi-input multi-output systems.
|
||||
Pascoal et al.~\cite{pascoal00_navig_system_desig_using_time} presented complementary filters which can adapt with time for navigation system capable of estimating position and velocity using GPS and SONAR sensors.
|
||||
|
||||
The noise characteristics of the super sensor are governed by the norms of the complementary filters.
|
||||
Therefore, the proper design of the complementary filters for sensor fusion is of immense importance.
|
||||
The design of complementary filters is a complex task as they need to tuned as per the specification of the sensor.
|
||||
In many applications, analytical formulas of first and second order complementary filters are used~\cite{corke04_inert_visual_sensin_system_small_auton_helic,jensen13_basic_uas}.
|
||||
These filters are easy to tune and simple to implement using an analog circuit~\cite{moore19_capac_instr_sensor_fusion_high_bandw_nanop,cite:yong16_high_speed_vertic_posit_stage}.
|
||||
However, these low order complementary filters are not optimal, and high order complementary filters can lead to better fusion~\cite{jensen13_basic_uas,shaw90_bandw_enhan_posit_measur_using_measur_accel}.
|
||||
|
||||
Several design techniques have been proposed to design higher order complementary filters.
|
||||
Pascoal~\cite{pascoal00_navig_system_desig_using_time} used linear matrix inequalities (LMIs)~\cite{boyd94_linear} for the design of time varying complementary filters.
|
||||
LMIs were also used by Hua et al.~\cite{hua04_polyp_fir_compl_filter_contr_system} to design finite impulse response (FIR) filters for the active vibration isolation system at LIGO.
|
||||
Plummer~\cite{plummer06_optim_compl_filter_their_applic_motion_measur} proposed an optimal design method using the $\mathcal{H}_{\infty}$ synthesis and weighting functions representing the measurement noise of the sensors.
|
||||
|
||||
Although various methods have been presented in the literature for the design of complementary filters, there is a lack of general and simple framework that allows to shape the norm of complementary filters.
|
||||
Such a method would prove to be very useful as the noise of the ``supper sensor'' and its dynamical characteristics depend on the norm of the filters.
|
||||
This paper presents such a framework based on the $\mathcal{H}_\infty$ norm minimization.
|
||||
The proposed method is quite general and can be easily extended to a case where more than two complementary filters needs to be designed.
|
||||
The organization of this paper is as follows.
|
||||
Section~\ref{sec:requirements} presents the design requirements of ideal complementary filters.
|
||||
It also demonstrates how the noise and robustness characteristics of the ``super sensor'' can be transformed into upper bounds on the norm of the complementary filters.
|
||||
The framework for the design of complementary filters is detailed in Section~\ref{sec:hinf_method}.
|
||||
This is followed by the application of the design method to complementary filter design for the active vibration isolation at LIGO in Section~\ref{sec:application_ligo}.
|
||||
Finally, concluding remarks are presented in Section~\ref{sec:conclusion}.
|
||||
|
||||
\section{Complementary filters requirements}\label{sec:requirements}
|
||||
|
||||
Complementary filters provides a framework for fusing signals from different sensors.
|
||||
As the effectiveness of the fusion depends on the proper design of the complementary filters, they are expected to fulfill certain requirements.
|
||||
These requirements are discussed in this section.
|
||||
|
||||
\subsection{Complementary characteristics}\label{sec:sensor_fusion}
|
||||
|
||||
Consider a case where two different sensors are used for measuring the same quantity, $x$ in different frequency range.
|
||||
The inherent dynamics of the sensors is represented by transfer functions \(G_1(s)\) and \(G_2(s)\).
|
||||
The two sensor also have uncorrelated noise characteristics given by \(n_1\) and \(n_2\).
|
||||
The signals from these two sensors are fused using complementary filters \(H_1(s)\) and \(H_2(s)\).
|
||||
The architecture of sensor fusion using complementary filters is shown in Figure~\ref{fig:fusion_super_sensor}.
|
||||
The resulting sensor, termed as ``super sensor'', can have larger bandwidth and better noise characteristics in comparison to the individual sensor.
|
||||
This means that the super sensor provides an estimate \(\hat{x}\) of \(x\) which can be more accurate over a larger frequency band than the outputs of the individual sensors.
|
||||
Based on Figure~\ref{fig:fusion_super_sensor}, the estimate of the physical quantity as measured by the super sensor can be written as
|
||||
\begin{equation}
|
||||
\label{eq:comp_filter_estimate}
|
||||
\hat{x} = \left(G_1 H_1 + G_2 H_2\right) x + H_1 n_1 + H_2 n_2
|
||||
\end{equation}
|
||||
|
||||
\begin{figure}
|
||||
\centering
|
||||
\includegraphics[scale=1]{figs/fusion_super_sensor.pdf}
|
||||
\caption{\label{fig:fusion_super_sensor}Block diagram of sensor fusion with complementary filters}
|
||||
\end{figure}
|
||||
|
||||
The complementary property of filters \(H_1(s)\) and \(H_2(s)\) simply implies that the summation of their transfer functions is equal to unity.
|
||||
That is, unity magnitude and zero phase at all frequencies (filters which satisfies only the magnitude condition are referred as ``magnitude complementary filter pair'').
|
||||
Therefore, a pair of strict complementary filter needs to satisfy the following condition.
|
||||
|
||||
\begin{equation}
|
||||
\label{eq:comp_filter}
|
||||
H_1(s) + H_2(s) = 1
|
||||
\end{equation}
|
||||
|
||||
\subsection{Noise characterization}\label{sec:noise_filtering}
|
||||
|
||||
In order to compute the noise characteristics associated with the estimate $\hat{x}$, it is first assumed that the dynamics of the individual sensors are perfect:
|
||||
\begin{equation}
|
||||
\label{eq:perfect_dynamics}
|
||||
G_1(s) = G_2(s) = 1
|
||||
\end{equation}
|
||||
|
||||
The output of the super sensor, \(\hat{x}\), based on the block diagram shown in Figure~\ref{fig:fusion_super_sensor} can be written as
|
||||
\begin{equation}
|
||||
\label{eq:estimate_perfect_dyn}
|
||||
\hat{x} = x + H_1 n_1 + H_2 n_2
|
||||
\end{equation}
|
||||
|
||||
The complementary are operating only on the noise component of the individual sensor.
|
||||
Thus, this sensor fusion architecture permits to filter the noise of both sensors without introducing any distortion in the physical quantity to be measured.
|
||||
The estimation error, \(\delta x\), of the super sensor can be written as
|
||||
\begin{equation}
|
||||
\label{eq:estimate_error}
|
||||
\delta x \triangleq \hat{x} - x = H_1 n_1 + H_2 n_2
|
||||
\end{equation}
|
||||
|
||||
The power spectral density (PSD) of the super sensor's estimation error is given by
|
||||
\begin{equation}
|
||||
\label{eq:noise_filtering_psd}
|
||||
\Phi_{\delta x} = \left|H_1\right|^2 \Phi_{n_1} + \left|H_2\right|^2 \Phi_{n_2}
|
||||
\end{equation}
|
||||
where, $\Phi_{\delta x}$ is the PSD of estimation error, $\Phi_{n_1}$ and $\Phi_{n_2}$ are the PSDs of the noise associated with the individual sensor.
|
||||
It can be seen that the estimation error's PSD depends on the PSD of the noise in individual sensor as well as the norm of the complementary filters.
|
||||
Therefore, by properly shaping the norm of the complementary filters, it is possible to minimize the noise of the super sensor noise.
|
||||
|
||||
\subsection{Robustness requirements}\label{sec:fusion_robustness}
|
||||
In the previous subsection, the inherent sensor dynamics were ignored.
|
||||
However in the real system, the sensor dynamics is not equal to unity.
|
||||
In such cases, the output of the sensor is normalized using a filter whose transfer function is equal to the inverse of the sensor dynamics.
|
||||
There are two major concerns in using inversion.
|
||||
First being the sensors may not have been calibrated properly and the actual sensor dynamics is not exactly compensated by the inverse filter.
|
||||
The second problem is that the inversion of sensor dynamics can result in an improper transfer function and hence may not be physically realizable.
|
||||
We here suppose that the sensor dynamics can be inverted using a proper and stable transfer function \(\hat{G}_i(s)\).
|
||||
However, we suppose there exists a normalization error since \(\hat{G}_i^{-1}(s) G_i(s) \neq 1\).
|
||||
This normalization error can be represented using frequency dependent multiplicative uncertainty (Figure~\ref{fig:sensor_fusion_dynamic_uncertainty}).
|
||||
In Figure~\ref{fig:sensor_fusion_dynamic_uncertainty}, \(\Delta_i(s)\) satisfies \(\|\Delta_i(s)\|_\infty \le 1\) and \(|w_i(s)|\) is a frequency dependent weighting function that represents the uncertainty corresponding to the normalization error.
|
||||
|
||||
\begin{figure}
|
||||
\centering
|
||||
\includegraphics[scale=1]{figs/sensor_fusion_dynamic_uncertainty.pdf}
|
||||
\caption{\label{fig:sensor_fusion_dynamic_uncertainty} Representation of normalization error in sensor fusion using multiplicative uncertainty}
|
||||
\end{figure}
|
||||
|
||||
Based on Figure~\ref{fig:sensor_fusion_dynamic_uncertainty}, the super sensor dynamics can be written as
|
||||
\begin{equation}
|
||||
\label{eq:super_sensor_dyn_uncertainty}
|
||||
\frac{\hat{x}}{x} = 1 + w_1(s) H_1(s) \Delta_1(s) + w_2(s) H_2(s) \Delta_2(s)
|
||||
\end{equation}
|
||||
|
||||
The dynamics of the super sensor now depends on the weighting functions ($w_1(s),w_2(s)$) and the complementary filters ($H_1(s),H_2(s)$).
|
||||
|
||||
The robust stability of the fusion can be studied graphically (refer Figure~\ref{fig:uncertainty_set_super_sensor}).
|
||||
The frequency response of the fusion output is plotted in a complex plane.
|
||||
The unity transfer function leads to a point $(1,0)$ located on the real axis.
|
||||
The uncertainty associated with first sensor at a particular frequency is represented by a circle with the center at (1,0) and radius $|w_1H_1|$.
|
||||
The uncertainty associated with the second is also represented using a circle centered at any point on the circle representing uncertainty associated with the first sensor and radius equal to $|w_2H_2|$.
|
||||
Therefore, the overall uncertainty of the fusion is represented with a circle centered at (1,0) and radius equal to $|w_1H_1|+|w_2H_2|$.
|
||||
The maximum phase difference that can result from the fusion is found by drawing a tangent from the origin to the uncertainty circle of super sensor.
|
||||
Mathematically, the maximum phase difference at frequency $\omega$ that can result from fusion is given by
|
||||
\begin{equation}
|
||||
\label{eq:max_phase_uncertainty}
|
||||
\Delta\phi(\omega) = \arcsin\left( |w_1(j\omega) H_1(j\omega)| + |w_2(j\omega) H_2(j\omega)| \right)
|
||||
\end{equation}
|
||||
|
||||
\begin{figure}
|
||||
\centering
|
||||
\includegraphics[scale=1]{figs/uncertainty_set_super_sensor.pdf}
|
||||
\caption{\label{fig:uncertainty_set_super_sensor}Robustness analysis of sensor fusion in the complex plane. The uncertainty associated with the super sensor dynamics are represented with a solid circle while those associated with individual sensors are represented with dashed circles.}
|
||||
\end{figure}
|
||||
|
||||
A constraint on the maximum phase difference at a given frequency can be incorporated in the design problem using the following equation:
|
||||
\begin{equation}
|
||||
\label{eq:max_uncertainty_super_sensor}
|
||||
\max_\omega \big( \left|w_1 H_1\right| + \left|w_2 H_2\right|\big) < \sin\left( \Delta \phi_\text{max} \right)
|
||||
\end{equation}
|
||||
where \(\Delta \phi_\text{max}\) is the maximum allowable phase difference.
|
||||
It can also be inferred from the above equation that the magnitude of the complementary filter (\(|H_i|\)) should be tuned to a smaller value at the frequencies where the magnitude of weighing transfer functions (\(|w_i|\)) representing sensor uncertainty is large.
|
||||
|
||||
\section{Design formulation using \(\mathcal{H}_\infty\) synthesis}\label{sec:hinf_method}
|
||||
|
||||
In this section, the shaping of complementary filters is expressed as an optimal $\mathcal{H}_{\infty}$ synthesis problem.
|
||||
The synthesis goal is to shape the frequency response of the filters such that they satisfy the design requirements presented in Section~\ref{sec:requirements}.
|
||||
|
||||
\subsection{Synthesis problem formulation}\label{sec:hinf_synthesis}
|
||||
|
||||
The first step is to formulate the filter design problem as a generalized plant-controller structure~\cite{boyd91_linear}.
|
||||
The generalized plant and controller structure for complementary filters design is shown in Figure~\ref{fig:h_infinity_robust_fusion}.
|
||||
In the figure, $P(s)$ is the generalized plant, $u$ is the ``control input'', $v$ is the ``measured output'' and $H_2(s)$ is the controller (filter) to be designed.
|
||||
The regulated outputs of the generalized plant, $z_1$ and $z_2$, are given by
|
||||
\begin{equation}
|
||||
\begin{split}
|
||||
z_1 &= W_1(s)(1-H_2(s)) w = W_1(s) H_1(s) w \text{ by defining } H_{1}(s) \triangleq 1 - H_{2}(s)\\
|
||||
z_2 &= W_2(s) H_2(s) w
|
||||
\end{split}
|
||||
\end{equation}
|
||||
where $w$ is the ``exogenous input'' to the plant, $W_1(s), W_2(s)$ are the weighting functions for shaping the complementary filters.
|
||||
|
||||
The dynamics of the generalized plant can be written as
|
||||
\begin{equation}
|
||||
\label{eq:generalized_plant}
|
||||
\begin{bmatrix} z_1 \\ z_2 \\ v \end{bmatrix} = P(s) \begin{bmatrix} w\\u \end{bmatrix}; \quad P(s) = \begin{bmatrix}W_1(s) & -W_1(s) \\ 0 & W_2(s) \\ 1 & 0 \end{bmatrix}
|
||||
\end{equation}
|
||||
|
||||
The weighting functions are chosen based on the specifications and requirements set for the complementary filters (discussed in Section~\ref{sec:hinf_weighting_func}).
|
||||
The objective of the optimization is to design a filter $H_2(s)$ such that the following conditions are satisfied
|
||||
\begin{equation}
|
||||
\label{eq:comp_filter_problem_form}
|
||||
\begin{split}
|
||||
\left| \frac{z_{1}}{w} \right| &= |1-H_2(s)| \le \frac{1}{|W_1(s)|} \\
|
||||
\left| \frac{z_{1}}{w} \right| &= |H_2(s)| \le \frac{1}{|W_2(s)|}
|
||||
\end{split}, \quad \forall \omega \in \mathbb{R}^{+}
|
||||
\end{equation}
|
||||
|
||||
\begin{figure}
|
||||
\centering
|
||||
\includegraphics[scale=1]{figs/h_infinity_robust_fusion.pdf}
|
||||
\caption{\label{fig:h_infinity_robust_fusion}Generalized plant controller structure for the design of complementary filters}
|
||||
\end{figure}
|
||||
|
||||
Based on Figure~\ref{fig:h_infinity_robust_fusion}, the $\mathcal{H}_{\infty}$ synthesis problem for the complementary filters can be stated as
|
||||
\begin{quote}
|
||||
Find a stable transfer function, $H_2(s)$, which takes measured output, $v$, as input and generates a control input, $u$, such that the $\mathcal{H}_\infty$ norm of the generalized plant from exogenous input, $w$, to the regulated output, ${[z_1,z_2]}^T$ is less than unity.
|
||||
\end{quote}
|
||||
Mathematically, the synthesis objective can be written as
|
||||
\begin{equation}
|
||||
\begin{split}
|
||||
&\left\|\begin{matrix} \left[1 - H_2(s)\right] W_1(s) \\ H_2(s) W_2(s) \end{matrix}\right\|_\infty \le 1 \\
|
||||
\Longleftrightarrow & \left\|\begin{matrix} H_1(s) W_1(s) \\ H_2(s) W_2(s) \end{matrix}\right\|_\infty \le 1; \quad H_1(s) \triangleq 1 - H_2(s)
|
||||
\end{split}
|
||||
\label{eq:hinf_syn_obj}
|
||||
\end{equation}
|
||||
The above optimization problem can be efficiently solved in \textsc{Matlab}~\cite{MATLAB2009} using Riccati formulae, linear matrix inequality based method or maximum entropy method.
|
||||
|
||||
|
||||
\subsection{Design of weighting functions}\label{sec:hinf_weighting_func}
|
||||
The choice of weighting function governs the shape of the designed complementary filters.
|
||||
Therefore, it is very important that the design specifications are appropriately transformed into the weighting functions.
|
||||
The choice of weighting functions is also constrained by the following factors
|
||||
\begin{enumerate}
|
||||
\item[(a)] Only proper and stable transfer functions can be used as weighting functions
|
||||
\item[(b)] As the order of the designed filter is equal to the sum of the orders of the weighting functions, the order of the weighting function needs to be reasonably small to ensure the physical implementation of the designed complementary filters.
|
||||
This also reduces the computational cost of the optimization problem.
|
||||
\item[(c)] The complementary property of the filter imposes a fundamental limitations on the weighting functions.
|
||||
The imposes a restriction that the magnitude of the filters $H_1(s)$ and $H_2(s)$ cannot be made small simultaneously at the same frequency.
|
||||
\end{enumerate}
|
||||
The specifications of the complementary filters are typically expressed using the following parameters --- low frequency gain, high frequency gain, slope (order of the filter) and the crossover frequency.
|
||||
We propose a weighting function that allows to translate the above requirements by setting simple parameters:
|
||||
\begin{equation}
|
||||
\label{eq:weight_formula}
|
||||
W(s) = \displaystyle\left( \frac{
|
||||
\hfill{} \displaystyle\frac{1}{\omega_0} \sqrt{\frac{1 - \left(\displaystyle\frac{G_0}{G_c}\right)^{\displaystyle\frac{2}{n}}}{1 - {\left(\displaystyle\frac{G_c}{G_\infty}\right)}^{\displaystyle\frac{2}{n}}}} s + \left(\frac{G_0}{G_c}\right)^{\displaystyle\frac{1}{n}}
|
||||
}{
|
||||
\left(\displaystyle\frac{1}{G_\infty}\right)^{\displaystyle\frac{1}{n}}\displaystyle \frac{1}{\omega_0} \sqrt{\displaystyle\frac{1 - \left(\displaystyle\frac{G_0}{G_c}\right)^{\displaystyle\frac{2}{n}}}{1 - \left(\displaystyle\frac{G_c}{G_\infty}\right)^{\displaystyle\frac{2}{n}}}} s + \left(\displaystyle\frac{1}{G_c}\right)^{\displaystyle\frac{1}{n}}
|
||||
}\right)^n
|
||||
\end{equation}
|
||||
where, \(G_0 = \lim_{\omega \to 0} |W(j\omega)|\) is the low frequency gain, \(G_\infty = \lim_{\omega \to \infty} |W(j\omega)|\) is the high frequency gain, $\omega_c$ is the crossover frequency, \(G_c = |W(j\omega_c)|\) is the absolute gain at the crossover frequency and $n$ is the order of the filter.
|
||||
As an illustration, the magnitude of the frequency response of the weighting function with the parameters \(G_0 = 0.001\), \(G_\infty = 10\), \(\omega_c = \SI{10}{Hz}\), \(G_c = 2\), \(n = 3\) and having high pass characteristics is shown in Figure~\ref{fig:weight_formula}.
|
||||
|
||||
\begin{figure}
|
||||
\centering
|
||||
\includegraphics[scale=1]{figs/weight_formula.pdf}
|
||||
\caption{\label{fig:weight_formula}Magnitude of the frequency response of the weighting function obtained using equation~\eqref{eq:weight_formula} with the parameters \(G_0 = 0.001\), \(G_\infty = 10\), \(\omega_c = \SI{10}{Hz}\), \(G_c = 2\), \(n = 3\)}
|
||||
\end{figure}
|
||||
|
||||
\subsection{Verification}\label{sec:hinf_example}
|
||||
The proposed methodology for the design of complementary filters is now applied on the following example.
|
||||
|
||||
\begin{example}
|
||||
Design complementary filters with the merging frequency around 10~Hz.
|
||||
The low pass filter should have DC gain of 0.001 and slope -2 above the merging frequency.
|
||||
The high pass filter should have a slope of 3 below the merging frequency and 0.001 high frequency gain.
|
||||
\end{example}
|
||||
|
||||
\noindent The first step is to design the weighting functions that translate the above requirements.
|
||||
To do so, equation~\eqref{eq:weight_formula} is used.
|
||||
The parameters corresponding to $W_1(s)$ and $W_2(s)$ are listed in Table~\ref{tab:weights_params}.
|
||||
The obtained transfer functions of the weighting functions are
|
||||
\begin{equation}
|
||||
\begin{split}
|
||||
W_1(s) &= \dfrac{1000 (s+34.55)^2}{(s+3455)^2}\\
|
||||
W_2(s) &= \dfrac{0.1 (s+87.43)^3}{(s+4.058)^3}
|
||||
\end{split}
|
||||
\end{equation}
|
||||
|
||||
Using these weighting functions, the generalized plant is evaluated using equation~\eqref{eq:generalized_plant}.
|
||||
The optimal complementary filters are obtained by solving the optimization problem given by equation~\eqref{eq:hinf_syn_obj}.
|
||||
The complementary filters obtained after optimization are
|
||||
\begin{equation}
|
||||
\begin{split}
|
||||
H_1(s) &= \frac{10^{-8} (s+6.6\times 10^9) (s+3450)^2 (s^2 + 49s + 895)}{(s+6.6e^4) (s^2 + 106 s + 3\times 10^3) (s^2 + 72s + 3580)}\\
|
||||
H_2(s) &= \frac{(s+6.6\times 10^4) (s+160) (s+4)^3}{(s+6.6\times 10^4) (s^2 + 106 s + 3\times 10^3) (s^2 + 72s + 3580)}
|
||||
\end{split}
|
||||
\end{equation}
|
||||
|
||||
The obtained complementary filters are of order 5 which corresponds to the sum of the orders of the weighting functions used.
|
||||
The frequency responses of the designed complementary filters are shown in Figure~\ref{fig:hinf_filters_results}.
|
||||
It can be seen that the designed filters fulfills all the design specifications and hence demonstrates the effectiveness of the designed methodology (more complex real life example is taken up in Section~\ref{sec:application_ligo}).
|
||||
|
||||
\begin{table}
|
||||
\caption{\label{tab:weights_params}Parameters used for \(W_1(s)\) and \(W_2(s)\)}
|
||||
\centering
|
||||
\begin{tabularx}{0.5\linewidth}{Xcc}
|
||||
\toprule
|
||||
Parameter & \(W_1(s)\) & \(W_2(s)\)\\
|
||||
\midrule
|
||||
\(n\) & \(2\) & \(3\)\\
|
||||
\(G_c\) & \(0.5\) & \(0.5\)\\
|
||||
\(G_\infty\) & \(1000\) & \(0.1\)\\
|
||||
\(G_0\) & \(0.1\) & \(1000\)\\
|
||||
\(\omega_c\) [\(\si{Hz}\)] & \(11\) & \(10\)\\
|
||||
\bottomrule
|
||||
\end{tabularx}
|
||||
\end{table}
|
||||
|
||||
\begin{figure}
|
||||
\centering
|
||||
\includegraphics[scale=1]{figs/hinf_filters_results.pdf}
|
||||
\caption{\label{fig:hinf_filters_results}Bode plot of the weighting functions and designed complementary filters}
|
||||
\end{figure}
|
||||
|
||||
\subsection{Extension to set of three complementary filters}\label{sec:hinf_three_comp_filters}
|
||||
In certain applications, more than two sensors are used to measure the same quantity and can be merged together to form a ``super sensor''.
|
||||
In such case, a set of three (or more) complementary filters is required.
|
||||
It is here shown that the proposed method can be generalized for the design of a set of arbitrary number of complementary filters.
|
||||
The control objective is now to design of a set of $n$ complementary filters ($H_i(s), i=1,\cdots,n$) which satisfy the following conditions
|
||||
\begin{equation}
|
||||
\label{eq:hinf_problem_gen}
|
||||
\begin{split}
|
||||
&\sum_{i=0}^n H_i(s) = 1 \\
|
||||
&\left| H_i(s) \right| < \frac{1}{\left| W_i(s) \right|}
|
||||
\end{split}
|
||||
\end{equation}
|
||||
Here, we extend the method to a case of three complementary filters.
|
||||
The generalized plant controller setup for this case is shown in Figure~\ref{fig:comp_filter_three_hinf}.
|
||||
The synthesis objective is to design filters $H_2(s)$ and $H_3(s)$ such that the $\mathcal{H}_\infty$ norm from exogenous input $w$ to regulated output vector $[z_1,z_2,z_3]^T$ is less than unity.
|
||||
That is,
|
||||
\begin{equation}
|
||||
\label{eq:hinf_syn_obj_three}
|
||||
\begin{split}
|
||||
&\left\| \begin{matrix} \left[1 - H_2(s) - H_3(s)\right] W_1(s) \\ H_2(s) W_2(s) \\ H_3(s) W_3(s) \end{matrix} \right\|_\infty \le 1\\
|
||||
\equiv &\left\| \begin{matrix} H_1(s) W_1(s) \\ H_2(s) W_2(s) \\ H_3(s) W_3(s) \end{matrix} \right\|_\infty \le 1; \quad H_1(s) \triangleq 1 - H_2(s) - H_3(s)
|
||||
\end{split}
|
||||
\end{equation}
|
||||
\begin{figure}
|
||||
\centering
|
||||
\includegraphics[scale=1]{figs/comp_filter_three_hinf.pdf}
|
||||
\caption{\label{fig:comp_filter_three_hinf}Generalized plant controller setup for designing a set of three complementary filters using \(\mathcal{H}_\infty\) synthesis}
|
||||
\end{figure}
|
||||
To validate this synthesis method, let's take an example where 3 sensors are merged together.
|
||||
The three sensors are working in the following three frequency range --- below 1~Hz, between 1Hz to 10Hz and above 10Hz.
|
||||
The weighting functions used for the synthesis are
|
||||
\begin{equation}
|
||||
\begin{split}
|
||||
W_1(s) &= \dfrac{1000 (s+3.141)^2}{(s+314.1)^2}\\
|
||||
W_2(s) &= \dfrac{2200 (s+62.83)^2 (s+6.283)^2}{(s+6283)^2 (s+0.06283)^2}\\
|
||||
W_3(s) &= \dfrac{0.1 (s+87.43)^3}{(s+4.058)^3}
|
||||
\end{split}
|
||||
\end{equation}
|
||||
The complementary filters are obtained by solving the optimization problem given by equation~\eqref{eq:hinf_syn_obj_three}.
|
||||
The frequency response of the designed filters and the weighting functions are shown in Figure~\ref{fig:three_complementary_filters_results}.
|
||||
|
||||
\begin{figure}
|
||||
\centering
|
||||
\includegraphics[scale=1]{figs/three_complementary_filters_results.pdf}
|
||||
\caption{\label{fig:three_complementary_filters_results}Bode plot of the weighting functions and designed set of three complementary filters}
|
||||
\end{figure}
|
||||
|
||||
\section{Application: Complementary Filter Design for Active Vibration Isolation of LIGO}\label{sec:application_ligo}
|
||||
Gravitational waves can help in detection various astrophysical events occurring in our universe.
|
||||
This can also pave a path to validate theories built around the existence of gravitational waves.
|
||||
However, the detection of these waves is an arduous task owing to the extraordinary small strain experienced by the earth due to gravitational waves.
|
||||
Various methods have been proposed for their detection, out of which laser interferometers are the most popular ones.
|
||||
Laser interferometers offers large projection range and high displacement sensitivity.
|
||||
Among the existing detector, Laser interferometer gravitation-wave observatory (LIGO) is the most sensitive operational detector.
|
||||
LIGO consists of two longs arms, referred as beam tubes, that are placed orthogonal to each other.
|
||||
The arms of the LIGO accommodates a Michleson interferometer with a cavity (Fabry-Perot).
|
||||
The mirrors at the extremity of the cavity serve as inertial test masses which responds to the strain induced due to the gravitational waves.
|
||||
The optics of the LIGO are suspended like a pendulum.
|
||||
The schematics of the LIGO are shown in Figure~\ref{ligo_schematics}.
|
||||
|
||||
\begin{figure}
|
||||
\centering
|
||||
\includegraphics[width=0.8\linewidth]{figs/ligo.png}
|
||||
\caption{\label{ligo_schematics}Schematics of LIGO}
|
||||
\end{figure}
|
||||
|
||||
The isolation of the terrestrial interferometers is necessary in order to isolate the motion of the suspended inertial masses from the seismic ground motion.
|
||||
The vibration isolation system attenuates the effect of disturbances on the motion of the suspended masses in the gravitational wave band.
|
||||
The other sources that can affect the sensitivity of the LIGO are thermal noise, human activities, tidal motion etc.
|
||||
The suspensions of the LIGO needs to serve two purpose --- isolation and alignment.
|
||||
The alignment is also important in order to ensure that the interferometer beam is targeted at the center of the suspended mass.
|
||||
The current vibration isolation system for LIGO has seven different stages.
|
||||
In the first stage, hydraulically actuated external pre-isolators are used for attenuating large ground motions.
|
||||
Two stages of active electromagnetic isolation system are placed next to pre-isolators.
|
||||
This is followed by four stages of pendulum based passive isolation system.
|
||||
|
||||
In the active isolation stage of LIGO, different sensors are used to sense the same physical signal in different frequency range.
|
||||
For example, seismometers are used to sense the position of the platform in the frequency band 0.5--10~Hz while geophones are employed above 10~Hz.
|
||||
The signals recorded from different sensor are fused using complementary filters~\cite{hua05_low_ligo,hua04_polyp_fir_compl_filter_contr_system}.
|
||||
The stringent requirements on these filters complicate their design.
|
||||
Hua~\cite{hua05_low_ligo} proposed complementary FIR filters which were synthesized using convex optimization.
|
||||
The designed FIR filters were found to be compliant with the design specifications.
|
||||
However, the order of the designed filter was very high, which limits its application to a practical system.
|
||||
In this section, we demonstrate the design of complementary filters with the same specification using the proposed method based on $\mathcal{H}_\infty$- synthesis.
|
||||
|
||||
\subsection{Design specifications}\label{sec:ligo_specifications}
|
||||
The design specification of the complementary filters (as listed out in~\cite{hua05_low_ligo}) are as follows:
|
||||
\begin{enumerate}
|
||||
\item[(a)] In the frequency range \(0\)-\(\SI{0.008}{Hz}\): the high pass filter's magnitude should be less than \(8 \times 10^{-4}\).
|
||||
\item[(b)] For frequency range \(\SI{0.008}{Hz}\)-\(\SI{0.04}{Hz}\): slope of the high pass filter is equal to three.
|
||||
\item[(c)] Between \(\SI{0.04}{Hz}\)-\(\SI{0.1}{Hz}\) frequency range: the high pass filter's magnitude should be less than \(3\).
|
||||
\item[(d)] For frequencies above \(\SI{0.1}{Hz}\): the low pass filter's magnitude should be less than \(0.045\).
|
||||
\end{enumerate}
|
||||
The specification of the complementary filters are shown graphically by dashed black lines in Figure~\ref{fig:ligo_weights}.
|
||||
|
||||
\subsection{Weighting Functions Design}\label{sec:ligo_weights}
|
||||
As the synthesis objective of the complementary filters is described by Eq.~\eqref{eq:hinf_problem_gen}, it is clear that the weighting functions should be chosen such that their inverse magnitude represent the maximum allowed norm of the complementary filters.
|
||||
This can be done manually using by combining poles and zeros or using useful formulas such as Eq.~\eqref{eq:weight_formula}.
|
||||
It is important to note that the order of the filters should be kept reasonable small in order to keep the computational cost of the optimization reasonable.
|
||||
This will also ensure that the designed filters are realizable in the physical world.
|
||||
The transfer function representing weights should also be stable and minimum phase.
|
||||
|
||||
The weighting function corresponding to the low pass filter, \(w_L(s)\), is here taken as Type I Chebyshev filter.
|
||||
The order of the weighting function for low pass filter is set as 20.
|
||||
|
||||
The weighting function for the high pass filter, \(w_H(s)\), is designed in such a way that its magnitude response is as close as possible to the design specifications.
|
||||
This was achieved using a combination of high-, low- and band-pass filters in the particular frequency band.
|
||||
The overall order of the weighting function for high pass filter is 7.
|
||||
|
||||
The magnitude responses of the inverse of the designed weighting functions and their comparison with the specifications are shown in Figure~\ref{fig:ligo_weights}.
|
||||
It can be seen that the inverse of the designed weights, shown in solid blue line for high pass filter and solid red line for low pass filter, are close to the specifications shown in black dotted line.
|
||||
|
||||
\begin{figure}
|
||||
\centering
|
||||
\includegraphics[scale=1]{figs/ligo_weights.pdf}
|
||||
\caption{\label{fig:ligo_weights}Specifications and weighting functions magnitudes}
|
||||
\end{figure}
|
||||
|
||||
\subsection{\(\mathcal{H}_\infty\) synthesis of complementary filters}\label{sec:ligo_results}
|
||||
The complementary filters are designed using \(\mathcal{H}_\infty\) synthesis based on the architecture shown in Figure~\ref{fig:h_infinity_robust_fusion}.
|
||||
The generalized plant is obtained by substituting the transfer functions of the designed weighting functions in equation~\eqref{eq:generalized_plant}, i.e., $W_1(s)=w_H(s)$ and $W_2(s)=w_L(s)$.
|
||||
The objective of the \(\mathcal{H}_\infty\) synthesis is given by equation~\eqref{eq:hinf_syn_obj}.
|
||||
The optimization problem in solved in \textsc{Matlab}~\cite{MATLAB2009} using Ricatti method.
|
||||
The frequency responses of the designed optimal complementary filters are shown in Figure~\ref{fig:comp_fir_ligo_hinf}.
|
||||
The order of the filters obtained using \(\mathcal{H}_\infty\) optimization is \(27\).
|
||||
|
||||
|
||||
Now, we compare the designed complementary filters with the FIR filters designed by Hua~\cite{hua05_low_ligo}.
|
||||
The transfer function of the FIR filter, $G(\omega)$, is the Fourier transform of its coefficients, $g(n)$.
|
||||
That is,
|
||||
\begin{equation}
|
||||
G(\omega) = \sum_n g(n)e^{-j2n\pi\omega}
|
||||
\end{equation}
|
||||
The FIR filter synthesis was formulated as a convex optimization problem.
|
||||
The objective of the optimization problem was to find the filter's coefficients such that their norm are below the specified upper bounds.
|
||||
The optimization problem was solved using SeDuMi~\cite{sturm99_using_sedum} and the obtained order for the FIR filters is \(512\).
|
||||
The bode plot of the FIR filters are shown with dotted lines in Figure~\ref{fig:comp_fir_ligo_hinf}.
|
||||
It can be seen that frequency responses of the designed complementary filters matches quite well with those of the FIR filters.
|
||||
The designed complementary filters are of much lower order and can be implemented with less computational cost can the FIR filters.
|
||||
The proposed methodology for the design of complementary filters can be effectively employed to obtain physically realizable filters.
|
||||
|
||||
\begin{figure}
|
||||
\centering
|
||||
\includegraphics[scale=1]{figs/comp_fir_ligo_hinf.pdf}
|
||||
\caption{\label{fig:comp_fir_ligo_hinf}Bode plot of the filters designed using $\mathcal{H}_\infty$ synthesis and FIR filters~\cite{hua05_low_ligo}}
|
||||
\end{figure}
|
||||
|
||||
\section{Concluding remarks}\label{sec:conclusion}
|
||||
The measurements from the sensors are reliable only within its bandwidth.
|
||||
The signals from different sensors are usually fused in order to measure a physical quantity over larger bandwidth.
|
||||
The sensor obtained after fusion is called as super sensor as it has superior noise characteristics and wider bandwidth.
|
||||
Complementary filters are used for the combining the signals from different sensors.
|
||||
A new framework based on $\mathcal{H}_\infty$ synthesis has been presented in this paper to aid the design of complementary filters.
|
||||
The method presented allows to shape the complementary filters based on the design specifications.
|
||||
The task of filter design is posed as an $\mathcal{H}_{\infty}$ synthesis problem.
|
||||
The design specifications of the systems are transformed in the form of weighting functions.
|
||||
These weighting functions are used in the optimization problem to constraint the filter response in a frequency band.
|
||||
The method has also been demonstrated for designing a set of three complementary filters.
|
||||
The design frame is general, simple to implement and can easily be extended to difference scenarios of sensor fusion.
|
||||
The effectiveness of the method is demonstrated for a real life application where complementary filters are designed for active vibration isolation of Laser Interferometer Gravitational-Wave Observatory (LIGO).
|
||||
The filters designed with the proposed method have been with compared with the finite impulse response (FIR) filters.
|
||||
It was found that the filters designed using $\mathcal{H}_\infty$ have lower order compared to FIR filters.
|
||||
The designed filters are physically realizable and have lesser computational cost compared to FIR filters.
|
||||
The proposed method can be effectively used to shape complementary filters based on design specifications.
|
||||
The method can be further be extended for the design of robust complementary filters with desired noise characteristics considering uncertainties in the sensor dynamics.
|
||||
This is the focus of our future research.
|
||||
|
||||
\section*{Acknowledgment}
|
||||
The authors would like to acknowledge the help received from the French Community of Belgium for funding the FRIA Grant of Thomas Dehaeze (Grant No.~FC 31597).
|
||||
|
||||
\bibliographystyle{elsarticle-num}
|
||||
\bibliography{ref}
|
||||
\end{document}
|
||||
436
mohit/ref.bib
436
mohit/ref.bib
@@ -1,436 +0,0 @@
|
||||
@article{collette15_sensor_fusion_method_high_perfor,
|
||||
author = {Collette, Christophe and Matichard, Fabrice},
|
||||
title = {Sensor Fusion Methods for High Performance Active Vibration
|
||||
Isolation Systems},
|
||||
journal = {Journal of Sound and Vibration},
|
||||
volume = 342,
|
||||
pages = {1-21},
|
||||
year = 2015,
|
||||
doi = {10.1016/j.jsv.2015.01.006},
|
||||
url = {https://doi.org/10.1016/j.jsv.2015.01.006},
|
||||
}
|
||||
|
||||
@phdthesis{hua05_low_ligo,
|
||||
author = {Hua, Wensheng},
|
||||
school = {stanford university},
|
||||
title = {Low frequency vibration isolation and alignment system for
|
||||
advanced LIGO},
|
||||
year = 2005,
|
||||
}
|
||||
|
||||
@inproceedings{hua04_polyp_fir_compl_filter_contr_system,
|
||||
author = {Hua, Wensheng and Debra, B. and Hardham, T. and Lantz, T.
|
||||
and Giaime, A.},
|
||||
title = {Polyphase FIR Complementary Filters for Control Systems},
|
||||
booktitle = {Proceedings of ASPE Spring Topical Meeting on Control of
|
||||
Precision Systems},
|
||||
year = 2004,
|
||||
pages = {109--114},
|
||||
}
|
||||
|
||||
@article{matichard15_seism_isolat_advan_ligo,
|
||||
author = {Matichard, F and Lantz, B and Mittleman, R and Mason, K and
|
||||
Kissel, J and others},
|
||||
title = {Seismic Isolation of Advanced Ligo: Review of Strategy,
|
||||
Instrumentation and Performance},
|
||||
journal = {Classical and Quantum Gravity},
|
||||
volume = 32,
|
||||
number = 18,
|
||||
pages = 185003,
|
||||
year = 2015,
|
||||
publisher = {IOP Publishing},
|
||||
}
|
||||
|
||||
@article{min15_compl_filter_desig_angle_estim,
|
||||
author = {Min, Hyung Gi and Jeung, Eun Tae},
|
||||
title = {Complementary Filter Design for Angle Estimation Using Mems
|
||||
Accelerometer and Gyroscope},
|
||||
journal = {Department of Control and Instrumentation, Changwon
|
||||
National University, Changwon, Korea},
|
||||
pages = {641--773},
|
||||
year = 2015,
|
||||
}
|
||||
|
||||
@article{corke04_inert_visual_sensin_system_small_auton_helic,
|
||||
author = {Corke, Peter},
|
||||
title = {An Inertial and Visual Sensing System for a Small
|
||||
Autonomous Helicopter},
|
||||
journal = {Journal of Robotic Systems},
|
||||
volume = 21,
|
||||
number = 2,
|
||||
pages = {43-51},
|
||||
year = 2004,
|
||||
}
|
||||
|
||||
@inproceedings{jensen13_basic_uas,
|
||||
author = {Austin Jensen and Cal Coopmans and YangQuan Chen},
|
||||
title = {Basics and guidelines of complementary filters for small
|
||||
UAS navigation},
|
||||
booktitle = {2013 International Conference on Unmanned Aircraft Systems
|
||||
(ICUAS)},
|
||||
year = 2013,
|
||||
month = 5,
|
||||
}
|
||||
|
||||
@inproceedings{pascoal99_navig_system_desig_using_time,
|
||||
author = {A. Pascoal and I. Kaminer and P. Oliveira},
|
||||
title = {Navigation System Design Using Time-Varying Complementary
|
||||
Filters},
|
||||
booktitle = {Guidance, Navigation, and Control Conference and Exhibit},
|
||||
year = 1999,
|
||||
}
|
||||
|
||||
@article{zimmermann92_high_bandw_orien_measur_contr,
|
||||
author = {M. Zimmermann and W. Sulzer},
|
||||
title = {High Bandwidth Orientation Measurement and Control Based on
|
||||
Complementary Filtering},
|
||||
journal = {Robot Control 1991},
|
||||
pages = {525-530},
|
||||
year = 1992,
|
||||
publisher = {Elsevier},
|
||||
series = {Robot Control 1991},
|
||||
}
|
||||
|
||||
@inproceedings{baerveldt97_low_cost_low_weigh_attit,
|
||||
author = {A.-J. Baerveldt and R. Klang},
|
||||
title = {A Low-Cost and Low-Weight Attitude Estimation System for an
|
||||
Autonomous Helicopter},
|
||||
booktitle = {Proceedings of IEEE International Conference on Intelligent
|
||||
Engineering Systems},
|
||||
year = 1997,
|
||||
month = {-},
|
||||
}
|
||||
|
||||
@article{shaw90_bandw_enhan_posit_measur_using_measur_accel,
|
||||
author = {F.R. Shaw and K. Srinivasan},
|
||||
title = {Bandwidth Enhancement of Position Measurements Using
|
||||
Measured Acceleration},
|
||||
journal = {Mechanical Systems and Signal Processing},
|
||||
volume = 4,
|
||||
number = 1,
|
||||
pages = {23-38},
|
||||
year = 1990,
|
||||
}
|
||||
|
||||
@article{brown72_integ_navig_system_kalman_filter,
|
||||
author = {R. G. Brown},
|
||||
title = {Integrated Navigation Systems and Kalman Filtering: a
|
||||
Perspective},
|
||||
journal = {Navigation},
|
||||
volume = 19,
|
||||
number = 4,
|
||||
pages = {355-362},
|
||||
year = 1972,
|
||||
}
|
||||
|
||||
@article{mahony08_nonlin_compl_filter_special_orthog_group,
|
||||
author = {Robert Mahony and Tarek Hamel and Jean-Michel Pflimlin},
|
||||
title = {Nonlinear Complementary Filters on the Special Orthogonal
|
||||
Group},
|
||||
journal = {IEEE Transactions on Automatic Control},
|
||||
volume = 53,
|
||||
number = 5,
|
||||
pages = {1203-1218},
|
||||
year = 2008,
|
||||
}
|
||||
|
||||
@article{sturm99_using_sedum,
|
||||
author = {Sturm, Jos F},
|
||||
title = {Using Sedumi 1.02, a Matlab Toolbox for Optimization Over
|
||||
Symmetric Cones},
|
||||
journal = {Optimization methods and software},
|
||||
volume = 11,
|
||||
number = {1-4},
|
||||
pages = {625--653},
|
||||
year = 1999,
|
||||
publisher = {Taylor \& Francis},
|
||||
}
|
||||
|
||||
@article{cao20_adapt_compl_filter_based_post,
|
||||
author = {Cao, Mingcong and Hu, Chuan and Wang, Jinxiang and Wang,
|
||||
Rongrong and Chen, Nan},
|
||||
title = {Adaptive Complementary Filter-Based Post-Impact Control for
|
||||
Independently-Actuated and Differentially-Steered Autonomous
|
||||
Vehicles},
|
||||
journal = {Mechanical Systems and Signal Processing},
|
||||
volume = 144,
|
||||
pages = 106852,
|
||||
year = 2020,
|
||||
publisher = {Elsevier},
|
||||
}
|
||||
|
||||
@article{tjepkema12_sensor_fusion_activ_vibrat_isolat_precis_equip,
|
||||
author = {Tjepkema, D and van Dijk, Johannes and Soemers, HMJR},
|
||||
title = {Sensor Fusion for Active Vibration Isolation in Precision
|
||||
Equipment},
|
||||
journal = {Journal of Sound and Vibration},
|
||||
volume = 331,
|
||||
number = 4,
|
||||
pages = {735--749},
|
||||
year = 2012,
|
||||
publisher = {Elsevier},
|
||||
}
|
||||
|
||||
@article{verma21_virtual_sensor_fusion_high_precis_contr,
|
||||
author = {Verma, Mohit and Dehaeze, Thomas and Zhao, Guoying and
|
||||
Watchi, Jennifer and Collette, Christophe},
|
||||
title = {Virtual Sensor Fusion for High Precision Control},
|
||||
journal = {Mechanical Systems and Signal Processing},
|
||||
volume = 150,
|
||||
pages = 107241,
|
||||
year = 2021,
|
||||
publisher = {Elsevier},
|
||||
}
|
||||
|
||||
@article{ma04_frequen_weigh_adapt_contr_simul,
|
||||
author = {Ma, Kougen and Ghasemi-Nejhad, Mehrdad N},
|
||||
title = {Frequency-Weighted Adaptive Control for Simultaneous
|
||||
Precision Positioning and Vibration Suppression of Smart
|
||||
Structures},
|
||||
journal = {Smart materials and structures},
|
||||
volume = 13,
|
||||
number = 5,
|
||||
pages = 1143,
|
||||
year = 2004,
|
||||
publisher = {IOP Publishing},
|
||||
}
|
||||
|
||||
@article{anderson53_instr_approac_system_steer_comput,
|
||||
author = {Anderson, WG and Fritze, EH},
|
||||
title = {Instrument Approach System Steering Computer},
|
||||
journal = {Proceedings of the IRE},
|
||||
volume = 41,
|
||||
number = 2,
|
||||
pages = {219--228},
|
||||
year = 1953,
|
||||
publisher = {IEEE},
|
||||
}
|
||||
|
||||
@article{shaw90_bandw_enhan_posit_measur_using_measur_accel,
|
||||
author = {Shaw, FR and Srinivasan, K},
|
||||
title = {Bandwidth Enhancement of Position Measurements Using
|
||||
Measured Acceleration},
|
||||
journal = {Mechanical Systems and Signal Processing},
|
||||
volume = 4,
|
||||
number = 1,
|
||||
pages = {23--38},
|
||||
year = 1990,
|
||||
publisher = {Elsevier},
|
||||
}
|
||||
|
||||
@inproceedings{baerveldt97,
|
||||
author = {Baerveldt, A-J and Klang, Robert},
|
||||
title = {A low-cost and low-weight attitude estimation system for an
|
||||
autonomous helicopter},
|
||||
booktitle = {Proceedings of IEEE International Conference on Intelligent
|
||||
Engineering Systems},
|
||||
year = 1997,
|
||||
pages = {391--395},
|
||||
organization = {IEEE},
|
||||
}
|
||||
|
||||
@inproceedings{roberts03_low,
|
||||
author = {Roberts, Jonathan M and Corke, Peter I and Buskey, Gregg},
|
||||
title = {Low-cost flight control system for a small autonomous
|
||||
helicopter},
|
||||
booktitle = {2003 IEEE International Conference on Robotics and
|
||||
Automation (Cat. No. 03CH37422)},
|
||||
year = 2003,
|
||||
volume = 1,
|
||||
pages = {546--551},
|
||||
organization = {IEEE},
|
||||
}
|
||||
|
||||
@inproceedings{bachmann03_desig_marg_dof,
|
||||
author = {Bachmann, Eric R and Yun, Xiaoping and McKinney, Doug and
|
||||
McGhee, Robert B and Zyda, Michael J},
|
||||
title = {Design and implementation of MARG sensors for 3-DOF
|
||||
orientation measurement of rigid bodies},
|
||||
booktitle = {2003 IEEE International Conference on Robotics and
|
||||
Automation (Cat. No. 03CH37422)},
|
||||
year = 2003,
|
||||
volume = 1,
|
||||
pages = {1171--1178},
|
||||
organization = {IEEE},
|
||||
}
|
||||
|
||||
@article{salcudean91_global_conver_angul_veloc_obser,
|
||||
author = {Salcudean, S},
|
||||
title = {A Globally Convergent Angular Velocity Observer for Rigid
|
||||
Body Motion},
|
||||
journal = {IEEE transactions on Automatic Control},
|
||||
volume = 36,
|
||||
number = 12,
|
||||
pages = {1493--1497},
|
||||
year = 1991,
|
||||
publisher = {IEEE},
|
||||
}
|
||||
|
||||
@article{mahony08_nonlin_compl_filter_special_orthog_group,
|
||||
author = {Mahony, Robert and Hamel, Tarek and Pflimlin, Jean-Michel},
|
||||
title = {Nonlinear Complementary Filters on the Special Orthogonal
|
||||
Group},
|
||||
journal = {IEEE Transactions on automatic control},
|
||||
volume = 53,
|
||||
number = 5,
|
||||
pages = {1203--1218},
|
||||
year = 2008,
|
||||
publisher = {IEEE},
|
||||
}
|
||||
|
||||
@article{pascoal00_navig_system_desig_using_time,
|
||||
author = {Pascoal, Antonio and Kaminer, Isaac and Oliveira, Paulo},
|
||||
title = {Navigation System Design Using Time-Varying Complementary
|
||||
Filters},
|
||||
journal = {IEEE Transactions on Aerospace and Electronic Systems},
|
||||
volume = 36,
|
||||
number = 4,
|
||||
pages = {1099--1114},
|
||||
year = 2000,
|
||||
publisher = {IEEE},
|
||||
}
|
||||
|
||||
@inproceedings{ryzhkov18_compl,
|
||||
author = {Ryzhkov, L},
|
||||
title = {Complementary filter design for attitude determination},
|
||||
booktitle = {2018 IEEE 5th International Conference on Methods and
|
||||
Systems of Navigation and Motion Control (MSNMC)},
|
||||
year = 2018,
|
||||
pages = {214--217},
|
||||
organization = {IEEE},
|
||||
}
|
||||
|
||||
@article{plummer06_optim_compl_filter_their_applic_motion_measur,
|
||||
author = {Plummer, AR},
|
||||
title = {Optimal Complementary Filters and Their Application in
|
||||
Motion Measurement},
|
||||
journal = {Proceedings of the Institution of Mechanical Engineers,
|
||||
Part I: Journal of Systems and Control Engineering},
|
||||
volume = 220,
|
||||
number = 6,
|
||||
pages = {489--507},
|
||||
year = 2006,
|
||||
publisher = {Sage Publications Sage UK: London, England},
|
||||
}
|
||||
|
||||
@article{odry18_kalman_filter_mobil_robot_attit_estim,
|
||||
author = {Odry, {\'A}kos and Fuller, Robert and Rudas, Imre J and
|
||||
Odry, P{\'e}ter},
|
||||
title = {Kalman Filter for Mobile-Robot Attitude Estimation: Novel
|
||||
Optimized and Adaptive Solutions},
|
||||
journal = {Mechanical systems and signal processing},
|
||||
volume = 110,
|
||||
pages = {569--589},
|
||||
year = 2018,
|
||||
publisher = {Elsevier},
|
||||
}
|
||||
|
||||
@article{ren19_integ_gnss_hub_motion_estim,
|
||||
author = {Ren, Zhengru and Skjetne, Roger and Jiang, Zhiyu and Gao,
|
||||
Zhen and Verma, Amrit Shankar},
|
||||
title = {Integrated Gnss/imu Hub Motion Estimator for Offshore Wind
|
||||
Turbine Blade Installation},
|
||||
journal = {Mechanical Systems and Signal Processing},
|
||||
volume = 123,
|
||||
pages = {222--243},
|
||||
year = 2019,
|
||||
publisher = {Elsevier},
|
||||
}
|
||||
|
||||
@article{faria19_sensor_fusion_rotat_motion_recon,
|
||||
author = {Faria, CT},
|
||||
title = {Sensor Fusion and Rotational Motion Reconstruction Via
|
||||
Nonlinear State-Observers},
|
||||
journal = {Mechanical Systems and Signal Processing},
|
||||
volume = 114,
|
||||
pages = {571--578},
|
||||
year = 2019,
|
||||
publisher = {Elsevier},
|
||||
}
|
||||
|
||||
@article{liu18_innov_infor_fusion_method_with,
|
||||
author = {Liu, Yahui and Fan, Xiaoqian and Lv, Chen and Wu, Jian and
|
||||
Li, Liang and Ding, Dawei},
|
||||
title = {An Innovative Information Fusion Method With Adaptive
|
||||
Kalman Filter for Integrated Ins/gps Navigation of Autonomous
|
||||
Vehicles},
|
||||
journal = {Mechanical Systems and Signal Processing},
|
||||
volume = 100,
|
||||
pages = {605--616},
|
||||
year = 2018,
|
||||
publisher = {Elsevier},
|
||||
}
|
||||
|
||||
@book{boyd94_linear,
|
||||
author = {Boyd, Stephen P and El Ghaoui, Laurent and Feron, Eric and
|
||||
Balakrishnan, Venkataramanan},
|
||||
title = {Linear matrix inequalities in system and control theory},
|
||||
year = 1994,
|
||||
publisher = {SIAM},
|
||||
volume = 15,
|
||||
}
|
||||
|
||||
@book{boyd91_linear,
|
||||
author = {Boyd, Stephen P and Barratt, Craig H},
|
||||
title = {Linear controller design: limits of performance},
|
||||
year = 1991,
|
||||
publisher = {Prentice Hall Englewood Cliffs, NJ},
|
||||
}
|
||||
|
||||
@book{MATLAB2009,
|
||||
year = {2020},
|
||||
author = {MATLAB},
|
||||
title = {version 9.9.0 (R2020b)},
|
||||
publisher = {The MathWorks Inc.},
|
||||
address = {Natick, Massachusetts}
|
||||
}
|
||||
|
||||
@article{abdel15_const_low_cost_gps_filter,
|
||||
author = {Abdel Hafez, Mamoun F and Saadeddin, Kamal and Jarrah,
|
||||
Mohammad Amin},
|
||||
title = {Constrained Low-Cost Gps/ins Filter With Encoder Bias
|
||||
Estimation for Ground Vehicles Applications},
|
||||
journal = {Mechanical Systems and Signal Processing},
|
||||
volume = 58,
|
||||
pages = {285--297},
|
||||
year = 2015,
|
||||
publisher = {Elsevier},
|
||||
}
|
||||
|
||||
@article{biondi17_attit_recov_from_featur_track,
|
||||
author = {Biondi, Gabriele and Mauro, S and Mohtar, T and Pastorelli,
|
||||
S and Sorli, M},
|
||||
title = {Attitude Recovery From Feature Tracking for Estimating
|
||||
Angular Rate of Non-Cooperative Spacecraft},
|
||||
journal = {Mechanical Systems and Signal Processing},
|
||||
volume = 83,
|
||||
pages = {321--336},
|
||||
year = 2017,
|
||||
publisher = {Elsevier},
|
||||
}
|
||||
|
||||
@article{brown72_integ_navig_system_kalman_filter,
|
||||
author = {R. G. Brown},
|
||||
title = {Integrated Navigation Systems and Kalman Filtering: a
|
||||
Perspective},
|
||||
journal = {Navigation},
|
||||
volume = 19,
|
||||
number = 4,
|
||||
pages = {355-362},
|
||||
year = 1972,
|
||||
doi = {10.1002/j.2161-4296.1972.tb01706.x},
|
||||
url = {https://doi.org/10.1002/j.2161-4296.1972.tb01706.x},
|
||||
}
|
||||
|
||||
@article{moore19_capac_instr_sensor_fusion_high_bandw_nanop,
|
||||
author = {Steven Ian Moore and Andrew J. Fleming and Yuen Kuan Yong},
|
||||
title = {Capacitive Instrumentation and Sensor Fusion for
|
||||
High-Bandwidth Nanopositioning},
|
||||
journal = {IEEE Sensors Letters},
|
||||
volume = 3,
|
||||
number = 8,
|
||||
pages = {1-3},
|
||||
year = 2019,
|
||||
doi = {10.1109/lsens.2019.2933065},
|
||||
url = {https://doi.org/10.1109/lsens.2019.2933065},
|
||||
}
|
||||
BIN
review/MV_dehaeze21_desig_compl_filte.pdf
Normal file
BIN
review/MV_dehaeze21_desig_compl_filte.pdf
Normal file
Binary file not shown.
BIN
review/dehaeze21_desig_compl_filte_cc.pdf
Normal file
BIN
review/dehaeze21_desig_compl_filte_cc.pdf
Normal file
Binary file not shown.
@@ -3,9 +3,9 @@
|
||||
"http://www.w3.org/TR/xhtml1/DTD/xhtml1-strict.dtd">
|
||||
<html xmlns="http://www.w3.org/1999/xhtml" lang="en" xml:lang="en">
|
||||
<head>
|
||||
<!-- 2021-09-01 mer. 10:15 -->
|
||||
<!-- 2021-09-02 jeu. 10:03 -->
|
||||
<meta http-equiv="Content-Type" content="text/html;charset=utf-8" />
|
||||
<title>A new method of designing complementary filters for sensor fusion using the $\mathcal{H}_\infty$ synthesis - Tikz Figures</title>
|
||||
<title>Designing complementary filters for sensor fusion using $\mathcal{H}_\infty$ synthesis - Tikz Figures</title>
|
||||
<meta name="author" content="Thomas Dehaeze" />
|
||||
<meta name="generator" content="Org Mode" />
|
||||
<link rel="stylesheet" type="text/css" href="https://research.tdehaeze.xyz/css/style.css"/>
|
||||
@@ -34,30 +34,30 @@
|
||||
|
|
||||
<a accesskey="H" href="../index.html"> HOME </a>
|
||||
</div><div id="content" class="content">
|
||||
<h1 class="title">A new method of designing complementary filters for sensor fusion using the \(\mathcal{H}_\infty\) synthesis - Tikz Figures</h1>
|
||||
<h1 class="title">Designing complementary filters for sensor fusion using \(\mathcal{H}_\infty\) synthesis - Tikz Figures</h1>
|
||||
<div id="table-of-contents" role="doc-toc">
|
||||
<h2>Table of Contents</h2>
|
||||
<div id="text-table-of-contents" role="doc-toc">
|
||||
<ul>
|
||||
<li><a href="#orgaddbc4b">1. Sensor Fusion - Overview</a></li>
|
||||
<li><a href="#orgcde34ed">2. Sensor Model</a></li>
|
||||
<li><a href="#org881c7cf">3. Sensor Model with calibration</a></li>
|
||||
<li><a href="#org52dba0e">4. Sensor Fusion Architecture</a></li>
|
||||
<li><a href="#org51c6b92">5. Sensor Model with Uncertainty</a></li>
|
||||
<li><a href="#orgd0c3320">6. Sensor Model with Uncertainty - Simplified</a></li>
|
||||
<li><a href="#orgd991dd2">7. Sensor fusion architecture with sensor dynamics uncertainty</a></li>
|
||||
<li><a href="#orgb45d1ac">8. Uncertainty set of the super sensor dynamics</a></li>
|
||||
<li><a href="#orgd194580">9. Generalized plant used for \(\mathcal{H}_\infty\) synthesis of complementary filters</a></li>
|
||||
<li><a href="#orgebe7ec5">10. Architecture used for \(\mathcal{H}_\infty\) synthesis of complementary filters</a></li>
|
||||
<li><a href="#orgf9a2f39">11. LIGO Sensor Fusion Architecture</a></li>
|
||||
<li><a href="#orga2611bb">12. Closed-Loop Complementary Filters</a></li>
|
||||
<li><a href="#org5af72e4">13. Closed-Loop Fusion Architecture</a></li>
|
||||
<li><a href="#org4e79ee7">14. Feedback Loop Sensor Fusion Architecture</a></li>
|
||||
<li><a href="#orgb59c88b">15. Feedback Sensor Fusion - Generalized Plant</a></li>
|
||||
<li><a href="#orge593f6d">16. Sensor Fusion - Sequential</a></li>
|
||||
<li><a href="#org3b323e8">17. Sensor Fusion - Parallel</a></li>
|
||||
<li><a href="#org9baa5b5">18. Architecture for \(\mathcal{H}_\infty\) synthesis of three complementary filters</a></li>
|
||||
<li><a href="#org10ffb09">19. Architecture for \(\mathcal{H}_\infty\) synthesis of three complementary filters</a></li>
|
||||
<li><a href="#org50515cc">1. Sensor Fusion - Overview</a></li>
|
||||
<li><a href="#orgd0f42e8">2. Sensor Model</a></li>
|
||||
<li><a href="#orgfe2b210">3. Sensor Model with calibration</a></li>
|
||||
<li><a href="#org7a71f40">4. Sensor Fusion Architecture</a></li>
|
||||
<li><a href="#org02ee7f2">5. Sensor Model with Uncertainty</a></li>
|
||||
<li><a href="#org15f5c37">6. Sensor Model with Uncertainty - Simplified</a></li>
|
||||
<li><a href="#org6703507">7. Sensor fusion architecture with sensor dynamics uncertainty</a></li>
|
||||
<li><a href="#org730783f">8. Uncertainty set of the super sensor dynamics</a></li>
|
||||
<li><a href="#org38b031f">9. Generalized plant used for \(\mathcal{H}_\infty\) synthesis of complementary filters</a></li>
|
||||
<li><a href="#orgc8963ca">10. Architecture used for \(\mathcal{H}_\infty\) synthesis of complementary filters</a></li>
|
||||
<li><a href="#orgc9e8b5c">11. LIGO Sensor Fusion Architecture</a></li>
|
||||
<li><a href="#org852a702">12. Closed-Loop Complementary Filters</a></li>
|
||||
<li><a href="#org21e2e13">13. Closed-Loop Fusion Architecture</a></li>
|
||||
<li><a href="#org784c1ac">14. Feedback Loop Sensor Fusion Architecture</a></li>
|
||||
<li><a href="#org08b29ce">15. Feedback Sensor Fusion - Generalized Plant</a></li>
|
||||
<li><a href="#org9bf391e">16. Sensor Fusion - Sequential</a></li>
|
||||
<li><a href="#orgc2ef6aa">17. Sensor Fusion - Parallel</a></li>
|
||||
<li><a href="#org90ea882">18. Architecture for \(\mathcal{H}_\infty\) synthesis of three complementary filters</a></li>
|
||||
<li><a href="#orge8d8968">19. Architecture for \(\mathcal{H}_\infty\) synthesis of three complementary filters</a></li>
|
||||
</ul>
|
||||
</div>
|
||||
</div>
|
||||
@@ -65,8 +65,8 @@
|
||||
Configuration file is accessible <a href="config.html">here</a>.
|
||||
</p>
|
||||
|
||||
<div id="outline-container-orgaddbc4b" class="outline-2">
|
||||
<h2 id="orgaddbc4b"><span class="section-number-2">1.</span> Sensor Fusion - Overview</h2>
|
||||
<div id="outline-container-org50515cc" class="outline-2">
|
||||
<h2 id="org50515cc"><span class="section-number-2">1.</span> Sensor Fusion - Overview</h2>
|
||||
<div class="outline-text-2" id="text-1">
|
||||
<div class="org-src-container">
|
||||
<pre class="src src-latex"><span class="org-font-latex-sedate">\definecolor</span>{myblue}{rgb}{0, 0.447, 0.741}
|
||||
@@ -106,7 +106,7 @@ Configuration file is accessible <a href="config.html">here</a>.
|
||||
</div>
|
||||
|
||||
|
||||
<div id="orgfdc79fa" class="figure">
|
||||
<div id="org3c4bfaf" class="figure">
|
||||
<p><img src="figs/sensor_fusion_overview.png" alt="sensor_fusion_overview.png" />
|
||||
</p>
|
||||
<p><span class="figure-number">Figure 1: </span>Schematic of a sensor fusion architecture using complementary filters</p>
|
||||
@@ -114,8 +114,8 @@ Configuration file is accessible <a href="config.html">here</a>.
|
||||
</div>
|
||||
</div>
|
||||
|
||||
<div id="outline-container-orgcde34ed" class="outline-2">
|
||||
<h2 id="orgcde34ed"><span class="section-number-2">2.</span> Sensor Model</h2>
|
||||
<div id="outline-container-orgd0f42e8" class="outline-2">
|
||||
<h2 id="orgd0f42e8"><span class="section-number-2">2.</span> Sensor Model</h2>
|
||||
<div class="outline-text-2" id="text-2">
|
||||
<div class="org-src-container">
|
||||
<pre class="src src-latex"><span class="org-font-latex-sedate"><span class="org-keyword">\begin</span></span>{<span class="org-function-name">tikzpicture</span>}
|
||||
@@ -136,7 +136,7 @@ Configuration file is accessible <a href="config.html">here</a>.
|
||||
</div>
|
||||
|
||||
|
||||
<div id="orga6c4cb8" class="figure">
|
||||
<div id="org9ced2bb" class="figure">
|
||||
<p><img src="figs/sensor_model.png" alt="sensor_model.png" />
|
||||
</p>
|
||||
<p><span class="figure-number">Figure 2: </span>Basic sensor model consisting of a noise input \(n_i\) and a linear time invariant transfer function \(G_i(s)\)</p>
|
||||
@@ -144,8 +144,8 @@ Configuration file is accessible <a href="config.html">here</a>.
|
||||
</div>
|
||||
</div>
|
||||
|
||||
<div id="outline-container-org881c7cf" class="outline-2">
|
||||
<h2 id="org881c7cf"><span class="section-number-2">3.</span> Sensor Model with calibration</h2>
|
||||
<div id="outline-container-orgfe2b210" class="outline-2">
|
||||
<h2 id="orgfe2b210"><span class="section-number-2">3.</span> Sensor Model with calibration</h2>
|
||||
<div class="outline-text-2" id="text-3">
|
||||
<div class="org-src-container">
|
||||
<pre class="src src-latex"><span class="org-font-latex-sedate"><span class="org-keyword">\begin</span></span>{<span class="org-function-name">tikzpicture</span>}
|
||||
@@ -171,7 +171,7 @@ Configuration file is accessible <a href="config.html">here</a>.
|
||||
</div>
|
||||
|
||||
|
||||
<div id="orgd38eff7" class="figure">
|
||||
<div id="org2edc54e" class="figure">
|
||||
<p><img src="figs/sensor_model_calibrated.png" alt="sensor_model_calibrated.png" />
|
||||
</p>
|
||||
<p><span class="figure-number">Figure 3: </span>Normalized sensors using the inverse of an estimate \(\hat{G}_i(s)\) of the sensor dynamics</p>
|
||||
@@ -179,8 +179,8 @@ Configuration file is accessible <a href="config.html">here</a>.
|
||||
</div>
|
||||
</div>
|
||||
|
||||
<div id="outline-container-org52dba0e" class="outline-2">
|
||||
<h2 id="org52dba0e"><span class="section-number-2">4.</span> Sensor Fusion Architecture</h2>
|
||||
<div id="outline-container-org7a71f40" class="outline-2">
|
||||
<h2 id="org7a71f40"><span class="section-number-2">4.</span> Sensor Fusion Architecture</h2>
|
||||
<div class="outline-text-2" id="text-4">
|
||||
<div class="org-src-container">
|
||||
<pre class="src src-latex"><span class="org-font-latex-sedate">\definecolor</span>{myblue}{rgb}{0, 0.447, 0.741}
|
||||
@@ -232,7 +232,7 @@ Configuration file is accessible <a href="config.html">here</a>.
|
||||
</div>
|
||||
|
||||
|
||||
<div id="org63acf85" class="figure">
|
||||
<div id="org895e3d1" class="figure">
|
||||
<p><img src="figs/fusion_super_sensor.png" alt="fusion_super_sensor.png" />
|
||||
</p>
|
||||
<p><span class="figure-number">Figure 4: </span>Sensor fusion architecture with two normalized sensors</p>
|
||||
@@ -240,8 +240,8 @@ Configuration file is accessible <a href="config.html">here</a>.
|
||||
</div>
|
||||
</div>
|
||||
|
||||
<div id="outline-container-org51c6b92" class="outline-2">
|
||||
<h2 id="org51c6b92"><span class="section-number-2">5.</span> Sensor Model with Uncertainty</h2>
|
||||
<div id="outline-container-org02ee7f2" class="outline-2">
|
||||
<h2 id="org02ee7f2"><span class="section-number-2">5.</span> Sensor Model with Uncertainty</h2>
|
||||
<div class="outline-text-2" id="text-5">
|
||||
<div class="org-src-container">
|
||||
<pre class="src src-latex"><span class="org-font-latex-sedate"><span class="org-keyword">\begin</span></span>{<span class="org-function-name">tikzpicture</span>}
|
||||
@@ -275,7 +275,7 @@ Configuration file is accessible <a href="config.html">here</a>.
|
||||
</div>
|
||||
|
||||
|
||||
<div id="org937c435" class="figure">
|
||||
<div id="orga300be8" class="figure">
|
||||
<p><img src="figs/sensor_model_uncertainty.png" alt="sensor_model_uncertainty.png" />
|
||||
</p>
|
||||
<p><span class="figure-number">Figure 5: </span>Sensor with multiplicative input uncertainty</p>
|
||||
@@ -283,8 +283,8 @@ Configuration file is accessible <a href="config.html">here</a>.
|
||||
</div>
|
||||
</div>
|
||||
|
||||
<div id="outline-container-orgd0c3320" class="outline-2">
|
||||
<h2 id="orgd0c3320"><span class="section-number-2">6.</span> Sensor Model with Uncertainty - Simplified</h2>
|
||||
<div id="outline-container-org15f5c37" class="outline-2">
|
||||
<h2 id="org15f5c37"><span class="section-number-2">6.</span> Sensor Model with Uncertainty - Simplified</h2>
|
||||
<div class="outline-text-2" id="text-6">
|
||||
<div class="org-src-container">
|
||||
<pre class="src src-latex"><span class="org-font-latex-sedate"><span class="org-keyword">\begin</span></span>{<span class="org-function-name">tikzpicture</span>}
|
||||
@@ -311,7 +311,7 @@ Configuration file is accessible <a href="config.html">here</a>.
|
||||
</div>
|
||||
|
||||
|
||||
<div id="orgce8343e" class="figure">
|
||||
<div id="org7ccc258" class="figure">
|
||||
<p><img src="figs/sensor_model_uncertainty_simplified.png" alt="sensor_model_uncertainty_simplified.png" />
|
||||
</p>
|
||||
<p><span class="figure-number">Figure 6: </span>Simplified sensor model</p>
|
||||
@@ -319,8 +319,8 @@ Configuration file is accessible <a href="config.html">here</a>.
|
||||
</div>
|
||||
</div>
|
||||
|
||||
<div id="outline-container-orgd991dd2" class="outline-2">
|
||||
<h2 id="orgd991dd2"><span class="section-number-2">7.</span> Sensor fusion architecture with sensor dynamics uncertainty</h2>
|
||||
<div id="outline-container-org6703507" class="outline-2">
|
||||
<h2 id="org6703507"><span class="section-number-2">7.</span> Sensor fusion architecture with sensor dynamics uncertainty</h2>
|
||||
<div class="outline-text-2" id="text-7">
|
||||
<div class="org-src-container">
|
||||
<pre class="src src-latex"><span class="org-font-latex-sedate">\definecolor</span>{myblue}{rgb}{0, 0.447, 0.741}
|
||||
@@ -380,7 +380,7 @@ Configuration file is accessible <a href="config.html">here</a>.
|
||||
</div>
|
||||
|
||||
|
||||
<div id="org5b69e5a" class="figure">
|
||||
<div id="orgc80e315" class="figure">
|
||||
<p><img src="figs/sensor_fusion_dynamic_uncertainty.png" alt="sensor_fusion_dynamic_uncertainty.png" />
|
||||
</p>
|
||||
<p><span class="figure-number">Figure 7: </span>Sensor fusion architecture with sensor dynamics uncertainty</p>
|
||||
@@ -388,8 +388,8 @@ Configuration file is accessible <a href="config.html">here</a>.
|
||||
</div>
|
||||
</div>
|
||||
|
||||
<div id="outline-container-orgb45d1ac" class="outline-2">
|
||||
<h2 id="orgb45d1ac"><span class="section-number-2">8.</span> Uncertainty set of the super sensor dynamics</h2>
|
||||
<div id="outline-container-org730783f" class="outline-2">
|
||||
<h2 id="org730783f"><span class="section-number-2">8.</span> Uncertainty set of the super sensor dynamics</h2>
|
||||
<div class="outline-text-2" id="text-8">
|
||||
<div class="org-src-container">
|
||||
<pre class="src src-latex"><span class="org-font-latex-sedate">\definecolor</span>{myblue}{rgb}{0, 0.447, 0.741}
|
||||
@@ -432,7 +432,7 @@ Configuration file is accessible <a href="config.html">here</a>.
|
||||
</div>
|
||||
|
||||
|
||||
<div id="orgc80ec14" class="figure">
|
||||
<div id="org6687171" class="figure">
|
||||
<p><img src="figs/uncertainty_set_super_sensor.png" alt="uncertainty_set_super_sensor.png" />
|
||||
</p>
|
||||
<p><span class="figure-number">Figure 8: </span>Uncertainty region of the super sensor dynamics in the complex plane (grey circle). The contribution of both sensors 1 and 2 to the total uncertainty are represented respectively by a blue circle and a red circle. The frequency dependency \(\omega\) is here omitted.</p>
|
||||
@@ -440,8 +440,8 @@ Configuration file is accessible <a href="config.html">here</a>.
|
||||
</div>
|
||||
</div>
|
||||
|
||||
<div id="outline-container-orgd194580" class="outline-2">
|
||||
<h2 id="orgd194580"><span class="section-number-2">9.</span> Generalized plant used for \(\mathcal{H}_\infty\) synthesis of complementary filters</h2>
|
||||
<div id="outline-container-org38b031f" class="outline-2">
|
||||
<h2 id="org38b031f"><span class="section-number-2">9.</span> Generalized plant used for \(\mathcal{H}_\infty\) synthesis of complementary filters</h2>
|
||||
<div class="outline-text-2" id="text-9">
|
||||
<div class="org-src-container">
|
||||
<pre class="src src-latex"><span class="org-font-latex-sedate"><span class="org-keyword">\begin</span></span>{<span class="org-function-name">tikzpicture</span>}
|
||||
@@ -473,7 +473,7 @@ Configuration file is accessible <a href="config.html">here</a>.
|
||||
</div>
|
||||
|
||||
|
||||
<div id="org725e946" class="figure">
|
||||
<div id="orge10d26f" class="figure">
|
||||
<p><img src="figs/h_infinity_robust_fusion_plant.png" alt="h_infinity_robust_fusion_plant.png" />
|
||||
</p>
|
||||
<p><span class="figure-number">Figure 9: </span>Generalized plant used for \(\mathcal{H}_\infty\) synthesis of complementary filters</p>
|
||||
@@ -481,8 +481,8 @@ Configuration file is accessible <a href="config.html">here</a>.
|
||||
</div>
|
||||
</div>
|
||||
|
||||
<div id="outline-container-orgebe7ec5" class="outline-2">
|
||||
<h2 id="orgebe7ec5"><span class="section-number-2">10.</span> Architecture used for \(\mathcal{H}_\infty\) synthesis of complementary filters</h2>
|
||||
<div id="outline-container-orgc8963ca" class="outline-2">
|
||||
<h2 id="orgc8963ca"><span class="section-number-2">10.</span> Architecture used for \(\mathcal{H}_\infty\) synthesis of complementary filters</h2>
|
||||
<div class="outline-text-2" id="text-10">
|
||||
<div class="org-src-container">
|
||||
<pre class="src src-latex"><span class="org-font-latex-sedate"><span class="org-keyword">\begin</span></span>{<span class="org-function-name">tikzpicture</span>}
|
||||
@@ -514,7 +514,7 @@ Configuration file is accessible <a href="config.html">here</a>.
|
||||
</div>
|
||||
|
||||
|
||||
<div id="org47d276f" class="figure">
|
||||
<div id="org0646c95" class="figure">
|
||||
<p><img src="figs/h_infinity_robust_fusion_fb.png" alt="h_infinity_robust_fusion_fb.png" />
|
||||
</p>
|
||||
<p><span class="figure-number">Figure 10: </span>Generalized plant with the synthesized filter</p>
|
||||
@@ -522,8 +522,8 @@ Configuration file is accessible <a href="config.html">here</a>.
|
||||
</div>
|
||||
</div>
|
||||
|
||||
<div id="outline-container-orgf9a2f39" class="outline-2">
|
||||
<h2 id="orgf9a2f39"><span class="section-number-2">11.</span> LIGO Sensor Fusion Architecture</h2>
|
||||
<div id="outline-container-orgc9e8b5c" class="outline-2">
|
||||
<h2 id="orgc9e8b5c"><span class="section-number-2">11.</span> LIGO Sensor Fusion Architecture</h2>
|
||||
<div class="outline-text-2" id="text-11">
|
||||
<div class="org-src-container">
|
||||
<pre class="src src-latex"><span class="org-font-latex-sedate">\definecolor</span>{myblue}{rgb}{0, 0.447, 0.741}
|
||||
@@ -570,7 +570,7 @@ Configuration file is accessible <a href="config.html">here</a>.
|
||||
</div>
|
||||
|
||||
|
||||
<div id="org81fe1cf" class="figure">
|
||||
<div id="org845a82e" class="figure">
|
||||
<p><img src="figs/ligo_super_sensor_architecture.png" alt="ligo_super_sensor_architecture.png" />
|
||||
</p>
|
||||
<p><span class="figure-number">Figure 11: </span>Simplified block diagram of the sensor blending strategy for the first stage at the LIGO</p>
|
||||
@@ -578,8 +578,8 @@ Configuration file is accessible <a href="config.html">here</a>.
|
||||
</div>
|
||||
</div>
|
||||
|
||||
<div id="outline-container-orga2611bb" class="outline-2">
|
||||
<h2 id="orga2611bb"><span class="section-number-2">12.</span> Closed-Loop Complementary Filters</h2>
|
||||
<div id="outline-container-org852a702" class="outline-2">
|
||||
<h2 id="org852a702"><span class="section-number-2">12.</span> Closed-Loop Complementary Filters</h2>
|
||||
<div class="outline-text-2" id="text-12">
|
||||
<div class="org-src-container">
|
||||
<pre class="src src-latex"><span class="org-font-latex-sedate"><span class="org-keyword">\begin</span></span>{<span class="org-function-name">tikzpicture</span>}
|
||||
@@ -602,7 +602,7 @@ Configuration file is accessible <a href="config.html">here</a>.
|
||||
</div>
|
||||
|
||||
|
||||
<div id="org57ca8ce" class="figure">
|
||||
<div id="org89c3b66" class="figure">
|
||||
<p><img src="figs/feedback_sensor_fusion.png" alt="feedback_sensor_fusion.png" />
|
||||
</p>
|
||||
<p><span class="figure-number">Figure 12: </span>“Closed-Loop” complementary filters</p>
|
||||
@@ -610,8 +610,8 @@ Configuration file is accessible <a href="config.html">here</a>.
|
||||
</div>
|
||||
</div>
|
||||
|
||||
<div id="outline-container-org5af72e4" class="outline-2">
|
||||
<h2 id="org5af72e4"><span class="section-number-2">13.</span> Closed-Loop Fusion Architecture</h2>
|
||||
<div id="outline-container-org21e2e13" class="outline-2">
|
||||
<h2 id="org21e2e13"><span class="section-number-2">13.</span> Closed-Loop Fusion Architecture</h2>
|
||||
<div class="outline-text-2" id="text-13">
|
||||
<div class="org-src-container">
|
||||
<pre class="src src-latex"><span class="org-font-latex-sedate"><span class="org-keyword">\begin</span></span>{<span class="org-function-name">tikzpicture</span>}
|
||||
@@ -644,7 +644,7 @@ Configuration file is accessible <a href="config.html">here</a>.
|
||||
</div>
|
||||
|
||||
|
||||
<div id="org8886809" class="figure">
|
||||
<div id="org7c31417" class="figure">
|
||||
<p><img src="figs/feedback_sensor_fusion_arch.png" alt="feedback_sensor_fusion_arch.png" />
|
||||
</p>
|
||||
<p><span class="figure-number">Figure 13: </span>Classical feedback architecture used for sensor fusion</p>
|
||||
@@ -652,8 +652,8 @@ Configuration file is accessible <a href="config.html">here</a>.
|
||||
</div>
|
||||
</div>
|
||||
|
||||
<div id="outline-container-org4e79ee7" class="outline-2">
|
||||
<h2 id="org4e79ee7"><span class="section-number-2">14.</span> Feedback Loop Sensor Fusion Architecture</h2>
|
||||
<div id="outline-container-org784c1ac" class="outline-2">
|
||||
<h2 id="org784c1ac"><span class="section-number-2">14.</span> Feedback Loop Sensor Fusion Architecture</h2>
|
||||
<div class="outline-text-2" id="text-14">
|
||||
<div class="org-src-container">
|
||||
<pre class="src src-latex"><span class="org-font-latex-sedate"><span class="org-keyword">\begin</span></span>{<span class="org-function-name">tikzpicture</span>}
|
||||
@@ -675,7 +675,7 @@ Configuration file is accessible <a href="config.html">here</a>.
|
||||
</div>
|
||||
|
||||
|
||||
<div id="org705140c" class="figure">
|
||||
<div id="org3aa7290" class="figure">
|
||||
<p><img src="figs/feedback_synthesis_architecture.png" alt="feedback_synthesis_architecture.png" />
|
||||
</p>
|
||||
<p><span class="figure-number">Figure 14: </span>Feedback architecture with included weights</p>
|
||||
@@ -683,8 +683,8 @@ Configuration file is accessible <a href="config.html">here</a>.
|
||||
</div>
|
||||
</div>
|
||||
|
||||
<div id="outline-container-orgb59c88b" class="outline-2">
|
||||
<h2 id="orgb59c88b"><span class="section-number-2">15.</span> Feedback Sensor Fusion - Generalized Plant</h2>
|
||||
<div id="outline-container-org08b29ce" class="outline-2">
|
||||
<h2 id="org08b29ce"><span class="section-number-2">15.</span> Feedback Sensor Fusion - Generalized Plant</h2>
|
||||
<div class="outline-text-2" id="text-15">
|
||||
<div class="org-src-container">
|
||||
<pre class="src src-latex"><span class="org-font-latex-sedate"><span class="org-keyword">\begin</span></span>{<span class="org-function-name">tikzpicture</span>}
|
||||
@@ -716,7 +716,7 @@ Configuration file is accessible <a href="config.html">here</a>.
|
||||
</div>
|
||||
|
||||
|
||||
<div id="org32b295e" class="figure">
|
||||
<div id="orgc836b02" class="figure">
|
||||
<p><img src="figs/feedback_synthesis_architecture_generalized_plant.png" alt="feedback_synthesis_architecture_generalized_plant.png" />
|
||||
</p>
|
||||
<p><span class="figure-number">Figure 15: </span>Generalized plant used for the \(\mathcal{H}_\infty\) synthesis of “closed-loop” complementary filters</p>
|
||||
@@ -724,8 +724,8 @@ Configuration file is accessible <a href="config.html">here</a>.
|
||||
</div>
|
||||
</div>
|
||||
|
||||
<div id="outline-container-orge593f6d" class="outline-2">
|
||||
<h2 id="orge593f6d"><span class="section-number-2">16.</span> Sensor Fusion - Sequential</h2>
|
||||
<div id="outline-container-org9bf391e" class="outline-2">
|
||||
<h2 id="org9bf391e"><span class="section-number-2">16.</span> Sensor Fusion - Sequential</h2>
|
||||
<div class="outline-text-2" id="text-16">
|
||||
<div class="org-src-container">
|
||||
<pre class="src src-latex"><span class="org-font-latex-sedate"><span class="org-keyword">\begin</span></span>{<span class="org-function-name">tikzpicture</span>}
|
||||
@@ -767,7 +767,7 @@ Configuration file is accessible <a href="config.html">here</a>.
|
||||
</div>
|
||||
|
||||
|
||||
<div id="org6fe58a4" class="figure">
|
||||
<div id="org187b8ba" class="figure">
|
||||
<p><img src="figs/sensor_fusion_three_sequential.png" alt="sensor_fusion_three_sequential.png" />
|
||||
</p>
|
||||
<p><span class="figure-number">Figure 16: </span>Sequential fusion</p>
|
||||
@@ -775,8 +775,8 @@ Configuration file is accessible <a href="config.html">here</a>.
|
||||
</div>
|
||||
</div>
|
||||
|
||||
<div id="outline-container-org3b323e8" class="outline-2">
|
||||
<h2 id="org3b323e8"><span class="section-number-2">17.</span> Sensor Fusion - Parallel</h2>
|
||||
<div id="outline-container-orgc2ef6aa" class="outline-2">
|
||||
<h2 id="orgc2ef6aa"><span class="section-number-2">17.</span> Sensor Fusion - Parallel</h2>
|
||||
<div class="outline-text-2" id="text-17">
|
||||
<div class="org-src-container">
|
||||
<pre class="src src-latex"><span class="org-font-latex-sedate"><span class="org-keyword">\begin</span></span>{<span class="org-function-name">tikzpicture</span>}
|
||||
@@ -813,7 +813,7 @@ Configuration file is accessible <a href="config.html">here</a>.
|
||||
</div>
|
||||
|
||||
|
||||
<div id="org6ec849a" class="figure">
|
||||
<div id="org72ff7bd" class="figure">
|
||||
<p><img src="figs/sensor_fusion_three_parallel.png" alt="sensor_fusion_three_parallel.png" />
|
||||
</p>
|
||||
<p><span class="figure-number">Figure 17: </span>Parallel fusion</p>
|
||||
@@ -821,8 +821,8 @@ Configuration file is accessible <a href="config.html">here</a>.
|
||||
</div>
|
||||
</div>
|
||||
|
||||
<div id="outline-container-org9baa5b5" class="outline-2">
|
||||
<h2 id="org9baa5b5"><span class="section-number-2">18.</span> Architecture for \(\mathcal{H}_\infty\) synthesis of three complementary filters</h2>
|
||||
<div id="outline-container-org90ea882" class="outline-2">
|
||||
<h2 id="org90ea882"><span class="section-number-2">18.</span> Architecture for \(\mathcal{H}_\infty\) synthesis of three complementary filters</h2>
|
||||
<div class="outline-text-2" id="text-18">
|
||||
<div class="org-src-container">
|
||||
<pre class="src src-latex"> <span class="org-font-latex-sedate"><span class="org-keyword">\begin</span></span>{<span class="org-function-name">tikzpicture</span>}
|
||||
@@ -865,15 +865,15 @@ Configuration file is accessible <a href="config.html">here</a>.
|
||||
</div>
|
||||
|
||||
|
||||
<div id="org7b3470a" class="figure">
|
||||
<div id="org7b7c9c7" class="figure">
|
||||
<p><img src="figs/comp_filter_three_hinf_gen_plant.png" alt="comp_filter_three_hinf_gen_plant.png" />
|
||||
</p>
|
||||
<p><span class="figure-number">Figure 18: </span>Generalized plant for the \(\mathcal{H}_\infty\) synthesis of three complementary filters</p>
|
||||
</div>
|
||||
</div>
|
||||
</div>
|
||||
<div id="outline-container-org10ffb09" class="outline-2">
|
||||
<h2 id="org10ffb09"><span class="section-number-2">19.</span> Architecture for \(\mathcal{H}_\infty\) synthesis of three complementary filters</h2>
|
||||
<div id="outline-container-orge8d8968" class="outline-2">
|
||||
<h2 id="orge8d8968"><span class="section-number-2">19.</span> Architecture for \(\mathcal{H}_\infty\) synthesis of three complementary filters</h2>
|
||||
<div class="outline-text-2" id="text-19">
|
||||
<div class="org-src-container">
|
||||
<pre class="src src-latex"> <span class="org-font-latex-sedate"><span class="org-keyword">\begin</span></span>{<span class="org-function-name">tikzpicture</span>}
|
||||
@@ -916,7 +916,7 @@ Configuration file is accessible <a href="config.html">here</a>.
|
||||
</div>
|
||||
|
||||
|
||||
<div id="org71495ac" class="figure">
|
||||
<div id="org4f6f8fa" class="figure">
|
||||
<p><img src="figs/comp_filter_three_hinf_fb.png" alt="comp_filter_three_hinf_fb.png" />
|
||||
</p>
|
||||
<p><span class="figure-number">Figure 19: </span>Generalized plant with the synthesized filter for the \(\mathcal{H}_\infty\) synthesis of three complementary filters</p>
|
||||
@@ -926,7 +926,7 @@ Configuration file is accessible <a href="config.html">here</a>.
|
||||
</div>
|
||||
<div id="postamble" class="status">
|
||||
<p class="author">Author: Thomas Dehaeze</p>
|
||||
<p class="date">Created: 2021-09-01 mer. 10:15</p>
|
||||
<p class="date">Created: 2021-09-02 jeu. 10:03</p>
|
||||
</div>
|
||||
</body>
|
||||
</html>
|
||||
|
||||
@@ -1,4 +1,4 @@
|
||||
#+TITLE: A new method of designing complementary filters for sensor fusion using the $\mathcal{H}_\infty$ synthesis - Tikz Figures
|
||||
#+TITLE: Designing complementary filters for sensor fusion using $\mathcal{H}_\infty$ synthesis - Tikz Figures
|
||||
:DRAWER:
|
||||
#+HTML_LINK_HOME: ../index.html
|
||||
#+HTML_LINK_UP: ../index.html
|
||||
|
||||
Reference in New Issue
Block a user