Add mohit's version
This commit is contained in:
File diff suppressed because it is too large
Load Diff
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
@@ -0,0 +1,951 @@
|
|||||||
|
%PDF-1.4
|
||||||
|
%ª«¬
|
||||||
|
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
|
||||||
|
xœÅ½ËŽ$;¶ 8?_á?P.#�O@ Û- �ºk&hÔ�+¡5zпßë±Íƒôˆ:Y‘åY·p+ÓWÒ�F#7÷{ý¯?ÒíÀÿý…ÿ“ú¼ý_û㸧\õéûŸðñÿú#ßç(¥ÖQôYüµ õö—Rïe¶™zÏ·¿´ó>ZO¹•Æ�éö?þ8nÿé�”ú¤yûÛ #UýéM:;þ@PøŸüë ÿõçìßü§ÿ¿÷ÿò!ŽŒ =ðãÿ-ÝÎÿŽïø¿ñ·ÿüG¾ý?¤ûYŽÜÇíÿÃìþË iôûqÎVòyÆ·�4±ÿóŸœVþò´r ÷ÿþéšÖöÑK¦u~}ZKÓg ¥]/tûè%Ó*_ŸVŸ÷–û‘Jj×´Ö�^2úõiM¼±2ÇYsº¦µ~ô’iµ/OëL'–¦¥: /qûè%Óê_ŸVî÷”Ƭ½ækZëG/™Öøú´ÎyÇ¡Ã6Ê×ÞÚ>zÉ´æ—§…™ÜûÑ°ŸÆ% ¶�^2t|}^í¸·ÞFÇ,®y ½f^_—óe`qJ�=µÇ¼Ö�^3¯¯ úšò=Ï‚êqEn ½f^_—ôõ<îX˜Zs¿ö×öÑkæõuQ_KÃâàﵕk^ëG¯™××e}ØLœA©�õZ?zͼ¾.ìkOw ½£ç^¯y ½f^_—öæûÙ!«øsž×öÑkæõuqßæ‰ÎÚژ׼Ö�^3¯¯Ëûžó½@%Ÿ½]×ÐöÑkôÔ¯Ëû^Ê}¶7uºö×öÑkæõuyßñ-ç8S ýÒ½¶�^3¯¯ËûÞÇ}ÖYz~�ëG¯™××å=4ø{©s}¬×úÑkæõuy?Žr?ëÙ±Ë &äúÑkæuÉû2¯y½O*=Mê Õø¸¬_vGçöÅ9œþI~½±í£×̪qVu@ˆã? Ÿ^÷ÍúÑkf5¾8«ñNãe2éë2\¿=�{®©•>ëµ}Ö�^`Õ]ˆ{QÎ{Á:öε¯—�^0±¯KqM¬¥{ÅmÒê9®‰ ½`b_ãšô•\hÊ �W¹~ô‚‰}]Žkb¸€Û€¢~ÎvMlýèû AΉ�ƒ–Mªi<&¶~ô‚‰}]s÷Äò½×2ÄÔcbËG/˜Ø×UwM,w|~%×ÇÄÖ�^0±¯ëîœXžË“ZêåšØöÑ &öuå]ëý>!´N˜×ÄÖ�^0±Ÿ“üÐîîÇ.Ƕ�^à8ý9ÉŸk»w‘k¿&¶~ô‚‰ýœäÏPgÒ.Ƕ�^0±Ÿ“ü¹@6@h�í}^`Z?'÷óÙ¡H”1Çñx‘ëG/˜ØÏÉý YzîRlûèû9¹Ÿæ|Öƶ�^0±Ÿ“û :ó“6¶}ô‚‰ýœÜOýxÖƶ�^0±Ÿ“û¸wžµ±í£Lìçä~ªç³6¶}ô‚ÈÔÏÉýTºT¯w)¶~ò‚iýœÔO%QñZ¥ØöÑ &ösR Ëò¬‹m ½`b?'÷ÓqJñ*8„×ÄÖ�^0±Ÿ“û£IïÂ*•˜×úÉ ¦õsR¿×gElýäÓú9™ßªt®:Û5õ“Lëç$~ ÏJØúÉ ¦õsò¾&)\å(×¹~ò‚iýœ´g,
|
||||||
|
VãñP¥—^ìÿ9Iñù¤~Ÿ¼`Z—¤ÿº?îò¤¾`ù'&qæóþpÞèÏ/˜ÈùÉùaæW öV~b JùCyÒŸ_0‘¯úÖÃCû*×ú±ÿjþ¡Gÿ=Cí»~WOûÇô—ŸÖÊ÷vϘÿ¼eñNàGÁ‹ß0¥ E.~Ôö°Åo˜Àcwj›œøòãHÇü®Ãñ ÏzÞ$ß>¡çPÖóùö =„¥6È·ÿü³ ©÷šŽã``ö›&ðƒôÝß0¥ ¥îþŽ)ý m÷wLé)»¿cJ?H×ý
|
||||||
|
SúQªîï˜ÒÒtÇ”~�¢û;¦ôCýæÛ§ô£Œáß1¥ $ ÿŽ)ý Oø7LéG)¿cJ?Èþ SúAbðï˜Òr‚Ç”~�ü;¦ôC»ëÛ§ô£ääß1¥ ä%ÿ†)ý(%ùwLéÙÈ¿cJ?HDþ SúAòï˜ÒÒ�Ô~”yü;¦ôäú~s{ó ø>ü|-ßænxªÌ}6ö¿}BÏ5¹Ï¦þ÷Oè©÷ÙÐÿþ =Õá>›ùß?¡§
|
||||||
|
Üg#ÿÛ'ô\{ûlâÿ„žªnŸ
|
||||||
|
üïŸÐS½í³yÿýú�_úÛ'ô\úûlÚÿ„žj~Ÿ
|
||||||
|
ûïŸÐS±ï³Yÿíz®ò}6ê¿BOå½Ï&ý÷Oè©®÷Ù ÿþ =ô>›óß?¡§JÞgcþû'ôƒøØ·O蹦øÙ�ÿþ =?›ñß>¡ç*âg#þû'ôT>ülÂÿ„žê†Ÿ
|
||||||
|
øïŸÐSÁð³ùþýzª~6Þ¿}BÏ%ÂϦû÷Oh‹Óøùó ÓYç-·û8Úí¯ÿŠùðKÿþœÎƒä¥¤šÇ-å4î�™! 3Ĥþå¯<z{ý_z´tÀع®¸ÎÛ¼ýõo·ÿöo1ƒôïnÿýö×ÿüÇø«Öàe¿ M¦Õ‘fÿä7óú›zp|Û¸w\é/}pè.¥´œñ¯pGÏ£äÏ ;ÖþWL!UÜ}ÐxË's8?™Ã?þþ'mÖÖ±kÿ®÷ÿ…ßüuïÿ ŸÄ/{ÿÿøþìý—êý§&›ÎEæ»6ÀW~ô×í€/Ìâ—m�/ÌáÏöÀg“øÇ÷@>þíYêÀ/ÿèuðÑΙç?»¾ò£xò–ðŸh |eù¿0 ºöR?Æ'«ÿÙ ò¼—ñ��@üü=ãOGÁ5žN<c˼ÔÏôa!f÷ÖÜÆ-<ã€^ô˜EýÒ üÂ�ÒÍ:¡ƒõO~´ýª mãžJí#ò£ýWýhxÉÇ(Ÿ-ïøU?zb§Hÿÿø›óWýf>ïÇìØJß¹ºó>g®õ³ý÷¿è7¬+\¯ç'?ù/¿è'^ç<ÛñÙiùß~Ño¬ öÙ¶ýßÑOVì ^�ùÙùü¿è7{½§1ûg?ù ÑO6)~rPÒ±ÿ毘¾²QÔOŠGYIv.úqží»«i®’®u
|
||||||
|
/©íúÂ$¢œk� 꺾0�G)×:…—Ôt}9U`{¿á ƒ±Â#ÍÈÂûF�OlÅm
|
||||||
|
ß¹ß+ ·)|ßV\«
|
||||||
|
·)|çV|w¾mïáwHņÀœ³�o—ŠO}çÖ™¼²Ý—¤ÃÞqn›Ò [Ï}eJO½æ¶)½°éÜW¦ôÔen›Ò ÛÍ}eJOýå¶)½°ÑÜW¦ôÔYn›Ò [Ì}iJ{O¹}J¯k.÷%Õhï&·Mé…må~J[ÛæòÍÚÚS3»u.¯ìj÷•)=µ±Û¦ôÂ~v_™ÒS»mJ/ìd÷•)=µ®Û¦ô v_™ÒSÓºmJ/ì^÷¥)míêö½¬oÝW&ôÔ¨n›Ñ ;Ö}iJ{‹º}J¯ëU÷6ä:“ïµ!Ÿºã3ye›¼¯Lé©/Þ6¥6ÈûÊ”ž:âmSzak¼¯Lé©Þ6¥6ÅûÊ”žºàmSza;¼¯Liï·Íèu�ð¾4¡½óÝ>£×µÀû’©ý]�ñ•Íï~ʳµÍå{=[Ï-÷¶¹¼°÷Þ¦ôÔlo�Ñ »î}aBOmöÖ ½°ßÞ&ôÔ`o�Ð ;í}aBOõÖ ½°ÇÞW&´7ÕÛ&ôºîz_ñ›míô6÷ÙËúê}É�÷w•ÄvÔûº‡}óœý†Sz/DüfûîÖÜfòÜšŸiìSú~·æ'û”¾ßù yÆ>¥ïwk~B›±OéûÝšŸfìSú~·æ'TOSúv·æ'$û”¾ß¹Eþö¹|säowknsù
|
||||||
|
nÍOØ9ö)}¿[ó^Ž}JßïÖü„‘cŸÒ÷»5?áâاôýnÍ�, O3ún·æ'üûŒ¾ßù óÆÓ”¾Ý¹ä#l3ùÞ|„ݹÍä7¸5?¡ûاôýnÍOˆ>ö)}¿[óŠ�}JßïÖü„ÜcŸÒ÷»5?Òzì3úv·æ'„ O3úv·æ'T û”¾ß¹eIísùÞ,©'·æ>—ïwk~äÙfôýnÍ�ì!Û„¾ßù‘7d›Ð÷»5?2†lú~·æG®�}BßîÖüÀ²»Ï¾Ûù‘ äÉ�÷;tÄóú<Üš_/“SKŒRÒLé–°ûT+ÖŽR?å{Ê™òA傹^²Hî ÿÉsÜ{š9}ò“ÿ\™äæ�ëýLí„Æû—“uy¦ÕI~aé¸÷Ìÿ|2‰®Vz›Ä<ð—£”£çó{ÞÿŸþâ÷¼þ?�Â7½ý?�ß½ü®Pz›CKí^ǵ|Ï«ÿ“ßûžÿ'ø¦×þ'3ø³—þOVF¯38O\#ý<�ẇ×Ðï•ý¤æ+_ûŸþb.Øh•uÈÿÀ‹ÿéeÿÓ)üÙÂZýs U¦ÖÞK¯í{ÖýÏ~ð{–ýÏfðåíþTŠþ—Æÿ“N<Î2�÷©ÿdR �šqÕÆ<�^?ŠÞ6ï9 -uÍ°%h\ãQ+ùÉ»øEÓ(øw#µ4>›Fþ¶iäŽÝ1êç«q~Û4Ø=cÖMøã4>«ºþ5Ó ûªŽc~¶å»&#`à(æôÙ,>ižðkf‘Õ£-÷ÏöÅgõþ¿è…Lˆ�R�ó³µø¤§Ã¯™Em÷û4—ÏÖ¢¿¨ºú¸—ÒÝÁ:Òÿ–c¹�Ùea=ß8O�J‰i÷™,ú‘Ÿ#Õ{[Û—Fªn²BOÍI_äü¹Yn3Ù�q[oÇW:ãî�À˜ç™kÅÿ süóݦ”[â§Yª=cõŽš¦Ú ¬Ø/ZÚ÷™¦§™¾×ˆaÐÁG‚dˆýÑæœÕPÉG ƒž)Ìõ~Î~ ràè§ ‰5jÇd/"c 6sê�õY:,�جgKÇ´1¾³ÌæåÄ*þt†¿˜˜r`PjŠÞ-±1�ì‡åf¯jbتØÒ3Í× ÙXûM…àxè=3Ƶ<JÍpºT)~Œ1õ¦BÇÊáOúaÄ \Øj.NÌ/¥
|
||||||
|
k|Û°þ傱>R å¸Æõcžñ�PÆQ€Å¸qàK²WeàŸÎ4è¾4vB– Øè “ž×8h3Ð…Mhãâ�®qódÿ?bçqÏ�Í�j<úÌ…:H`|ç�çÁXme(Ã.á¯�ý G‰ïœ§O~C
|
||||||
|
Œ³ã(âá…aåczž'[€ Gg ¬V˜˜30¶hv�Ã[>OϳÜKë¥� ãðÒ:@XÅkÁ�1rŒKx{ø4°�I·‘®qÐ}±é…á–i'Þmò»ÍùÄLKÖï¹e¬ËÑk–£Æ
|
||||||
|
,Æ�ÇÉ“IhÜ[�
|
||||||
|
ÖELå,XÆᯜw<éYÕ7SÎFÂ
|
||||||
|
Mí°Wð¦1+ë•bºÆ¸RÙ`WËRÒ½T Æ\¯q³çê%£R[O «�À”¾Ã[åm�P± ñôÂN˜x-˜Ÿ±†xN-u)ü‰ ,ƵcÆKÍ>©Œ\³Ý‰0< @mk0œ ¬KŽq½ŒëõÁl9KeÿûkÜÄמžg¿Ókƒ½ØÀ3`‡ÖKîÀâ;yºg¬ç€LÄ“´ ãfærz.r‚û7°Žã1Â)ÁÞl †�4÷‰®Ç=0†ãŽM/Œo+ÑjŒƒtKÉ[Ø#¯È¥ì!áR¡iŸ'ž5^щƒ’1yaPú`);ëøÊéï„8ÇÙÇHË9þãÆ\
|
||||||
|
Ô4œÛvÆ8LzÔâgÀ=©‡©Æ8ümVK *Vø œÔ˜KøIË\`#ãvãÝ%¬B”Å#ЖÆ<â
|
||||||
|
Áx,Ü<ÂÏTæÅc¬Õ‚of ßc* .Gaäh8ðß ãz»r" ѱ‚ó ã01Ibé~@`‹qƒ„†Ä DtÈ›7!L »…±åpÅuW=®@æöxµ¸mŽŽfLMÉÏÀ ;ÆQãzÃY¯E±‚Ôʽ&ì™·[aㇳxX¥Ñ‹_‹Û�J/ÞœÀx¹ 5n7z *ý ÄÚ½ ½‘-ÂXa¼oëØ�íãðts瀖�ûºŒ‡·§ž€Æ:¤°w 6bO›Ë�¥ Ruú4tÒÀ§¬ƒ@‚æsa êÀQZŒØs3é;;þ
|
||||||
|
Máàe.l �ÀÞ� š n§˜Ëìn”m ;uàBð3`eûi¤C^Q‘Â]aGF�‡¦ ð˜Æ°Œ
|
||||||
|
tý äÕÔ½UƒŠå�Ô!®&v÷£ ˆêÇÓ±?TüyÄ°Òò¬Öw:ÄÕøoèI·3äx
|
||||||
|
Œ];�Å8 P,„ÞP lч÷3b\Ï ÷† �émPTƈq°‰)MƒÐ|›ÆæqàÎÒ3 È+(� ²'°™˜r)ŒÒò€ì,þNˆ£Rã†6±ùûyz\Kž3!ˆ«Þ[ç�,(sÛZ$�“ǯ$Ç0\Ï=¥QÔ÷ XŒƒŠ(£ÚXÇ›ì8ÔÆp
|
||||||
|
BéðT ®`é‹qV9ôèÐøbè
|
||||||
|
˜W½DËp
|
||||||
|
E‡H4†W‚�¤Í2 ®ð
|
||||||
|
!Xü¯â¹!¬ðý�� �¬>& a…{½öص#54¨Aî
|
||||||
|
h}½ú0#l o± a…×\y[ à i¢ß›VqµÇÏAD'_Î@pÊq)•€påè€cçO|{ ;¨|�Þ|’Êo<†MH üD`cyn¼ûS*!H*þ§‡ò;0?ì(«ŒC9°ìØ1¸üG@¸ê±$)†AT`cú+!¨Ør3” QOÈ»XÈ)ȶçq`ép{x”ˆpØ03¾‘§?ç¶th_2#nTA“·}ó0Üóg’6�¹l™.�¶Æhý‚¬»‡â;Ý¿
|
||||||
|
ã©Îò€ Â)åý#ˆÅ0
|
||||||
|
Rè½³qÊlžº¡÷¾‹¯Ì¦Íøi\1 ßY[ò¯ÞÑT“ ñ#Q Æ`4ž�ø)»mN7NŒn^bE›/Æ^´Ý¿'?3žaªãºEp©Âèƒú\‡Óé,,Tò30-Àâ+iQÏflPBBÓìžJ¢9¿6qÏ�œýÂéÇ�à›g ê vM‹a8¡ðg:p à¿1» €Æ%j<…FxŒ£qsJ\fRÑSW¯5ÆaP9k
|
||||||
|
³ã~5c 'S)žxû¼‡ùfíò€vŽ».+¸çqIÕÓ•øÃèU7 1Úkø(ÆaÝÈaUÝesÍ1nrqãù ë’=SK¹0,..˜éyâzÁTrK oü˜Ñ¬ƒ qd¯ ,qÞ
|
||||||
|
3^´»ÖmÐdšâk5Ä(í®30hXø{÷\|£x]hŠ aE ƒ¾C»Yl;6pž-Æ5ì¹CF1Úªø§1nà*ìºb2Mq¼h}
|
||||||
|
1ü©4 �LSü°%f /2†K 4šØàý‡—¦/1H¿_9húìÁÏú9<¬A ;�æáqí¤ê2ý�¾nÜû3ÆAæC ÷T:µDX>3Æ�|F*ŠˆÖ`ECˆÂ“ƒwj�ÁL_Ü€§ 0vú„À—×Á;$É°£išÓ1ß9XTç“ lP'=¥öfîGn4cT
|
||||||
|
p=�R{�ñ A ÃŽÄoͳƸN7‹Ÿï¤‹ ÿî¬ ‡yÕd¹sÒûŽc #À Ëd ò
|
||||||
|
Š ]Æ O !\N:¹öKñ•S–A ×o,~çIY†U—Ú‹ÛèNñ®§iÕ‹¥1ÌíÄ;®Ï7 7
|
||||||
|
±AS²O�KX|h :E°ÅyÒ+b¬6[LÄ ±:^p 1n`ƒtŸ>`<7½w�Öœ’Ä`‹7D¼´À°ðØiš‹’•`‰äâïÄ¢ãæëÆ2›¶ã2>=Žj;mE áô¸“ „f«©‰¸‡aãˆYÐÁ܆Û%{Ídè90nÇf¿ÛÉ^Á6îq×ÞéRlöº�”]£7‹eMšû U¹`‡N9Žp,Š•!ïwâ8687ÝßÙØɶÇÃA\Uúü¤õžÔÓÔOÄU… MÒzOØgЇëh�a§vÞvÆ ¶×Ã7,q^ê-u�æ‘W@ ‚ÀxA$ dÖG`:6WÔØ´-' âD>ÆÓßÉCAYÞ%ÞO•1\Îí¸ �N �¬Ðg†—†8l÷Ƒƺùµ„UÚÞ¸~’ÇÁì† ‘“%6¸ð!b“ÇÁ¶=ž Ý©ÓH„[jzòW*xX(d�5\5É ¡�KgÁéeÕ^�ç ¸ñ+µÿj oÆ¢GMHæ‚»Ç?†·Mé(V8¶+vÓð0ÙÒg¿° m^Ä z�WÞhà�WmÚÂÚé;–y,89ÿ éRÄ(.p?ÖX“ò/0˜ÐCcQúÙz2Ãñ6îVäbQ:ŒÄPÉ`iS�Kez¶F;¬nÃK„&<= & „ŒE ¬pˆPLdxÜ$Í]¶+<<8 Y±?ì?"6)ÞmMWÒbbãZ¬4¶�Çf/R}+½ò ÂóœJ•Iº M\�Õ‚ †öáÝîaxcÓóÂúã�c·AïÁ»Ô¯Ñ—
|
||||||
|
[= Ò|bõä´ÁùŸb¨à �`<T;J`8Žå�A" ‹âçæ�ýøFlEl}:ㄵ3_®Fb“~«"Õ—Fwá›°ÊŒ-¬Iö8쓇ÂI ‡>Y�àËÂû¦‹žlp\¡ôWžr臽fçsõ;˜ø±™cˆ"û”Ö/�¶ámÔãÀ½%ËŠž}ìÙã˜uèMä†�Ñ6!6âÀÎ9ý�Ð뜀Fˆ.s Ï3¾2huÉïgÐÀÉTwáf‘‚¿b
|
||||||
|
'kœÙÃð)$õàƒW~ZJ*ïÓéÅ34aÄã7< ’‚دnð4Ééž„áÑ
|
||||||
|
MaÜ ¸2OE*€á¸ Í*Âà:T9<�.âÀ°"И§ÇA%høÐsa„÷Bž ‡ï?°Ou”´SyxܤÜMÖþÆ�EÌãl z�
|
||||||
|
F8`keÙ&òwhfZMÖáC|»ä¡KèN“‹œØ„u M§Ã—·��
|
||||||
|
|
||||||
|
8da˜TÁÃßI½ût¤°6Ù \¯a2‘iDyœT¿2/ ÷ Ö¥x VsB»öïQŠN.�Ças4iÝÄá.¶Fc²+ClÐ)’†ŸÇ²RÆ$}Šu© èïä´! S¬ 6UŠ×>»¤šü„Z¹¨ H*üUŽ$A¥]5H \ ÐùS�aP¨p™ðÍž 8ë©Æ8?i –V¯–U8¢NS
|
||||||
|
ö®úÆÛÞ‹Òqsæbˆá
|
||||||
|
~ßéµÄû ÷£Õ krÆ°“RGB“üIøŒÌ�q—ïë„)Žá¯ŸãU+•‘v%Å8œÈ ÿ ÃPÁSò8,lÚÃO0ôÐtx\Ã�©`)1nUœ§$µX¦K<~�á�‚�«‚ÂM Ã'mÙxw4Mgõº0¼�ÍzHíå^)PÉG
|
||||||
|
ŒÞƒj‡<Í›³‡›·=¢ ò�Æ*$ªÎD®{ìh;ä»Tb…�Áò˜Õyœ!æ$wNf�ÕŠ›Î%Z«Ä¤>É<xH·UNãÞ!Þ¹Ø$[ ”HZÁí¸Ïç%`B†û
|
||||||
|
¯‰ ú-I5Ðct’ó8=.Ó-› ¤.Áaɉ7 ÓàÊâN- ¯«â„Ú€ 0µ"•
|
||||||
|
‰�ïÆðgšÍÃ
|
||||||
|
ì[Ø+~j†6 Íæq™\úpESXàŸT+N°�.w>_zGSê£�C7 1¬8ÄAñ8ú¦í>E¦Ø ãŠuÐ ÏŒl4Âa“tþyJO#†‹ K*/NjŽó�4ÃöóÌV•�öh�4Ó61Ç1� 7øÑ‚¬Ãª$û¤ðz³ú3ë=bQÂÏ‹›
|
||||||
|
2 yΛ¸E –”� �ðõ劇:‰áíE#ü¨Œ&pÜÁø9ª €q
|
||||||
|
èÊC‹¢@FáÄôWü £ÿü‡�Œî“Eü° JXÂ�·éS®�#,Tz3-χFbt¿A¥—sÆ›�uilLú[i©1ø�qÎJ°•Õʾq‰p>.ߟô{Ll0?ŽØÃ%d z nJ®
|
||||||
|
V�76ãɨceò Oa��yÂð5¸ÆâKÆñ©ô”ë4»S#‰A–ùjMYHßD§ÒËM…ÓÁáGg`#ômƦ»ÐØ|ÏRÐ VvCƒ¦šÚV$™Îݡȳ0¬IOZ'†¬[,Ü�Æ×àm§ƒñ2-õM9ª(iq@
|
||||||
|
éÅ, kL2D
|
||||||
|
aÌc€ÕÆ5ð{�†Ä—kõ *´éºÜ”!x4G]
|
||||||
|
A”aM¨Þ%*¨Øan•B ‘É>(9˜„Í€¥õÃ1¨�)tê¼uy¹·€1d 2_žL:˜ädÅ7Ü™= œãð8‰.<%ÃÐ&©aQV3�ä¸ œa
|
||||||
|
èT´l(ÿy»¥úÆ1(ÑÁY,Ã*£¸^µ;ú~G¾B6xx[åi2o)†�§·:‹Ý`Ó1åâÆÛ
|
||||||
|
:šeåÖ€ÑA©ˆ &�lØáô 'º‰o¼hqíT2Ÿ{‚›÷3tÇP™ Å3ºÊÉN,1Ø’|
|
||||||
|
*M¥p; tð`qÒÈAcŸJ»¢�7É3â™0¬!™ã¨™0ÿ f¸l˪5Á …ª Á lÖ;¯øÊ =•$¨•‘ùGf¦©^ÏCñ ¹ÄØLã0:ºf†a¦q��
|
||||||
|
˜¼tÆSË›Ý÷Y`ƒ†9½ñÄŽ^˜+ ˆ�
|
||||||
|
|#�ñ¡TÒ+%©ØÙ�¿=«? .©Ë‹C ÆÊd”èF%ï²—*G/v(tŸƒo‡j1—kôâod`V:ýxT§!¯DV-Œ Ô娅7šáG>¢ ŸÅ4Åü¿ƒ^¯ágc\Ò…´E0 räž Â°ÍZ®È`Û�>¥ ƒUD+ÂTA* Ù禷“d¸$´i°8èÆæ=R‚8Ë»<gY[Äpª¦‚W´¡xûÂ"ÑÃuÆ5èÙ%oÍ2
|
||||||
|
Ää·Ú×À©†v£9‡-C¿²ž®3®Q±(…ãü¯=Oo°Î¸þ?ó hXRˮγP_/Ì’¯ïFƒ´ÒèôEÞé[‚ZZÈÄ·õÊ\˜ ˜” ªá2�©NÛÑ 1ww~�º�“ï®ê’&S61VN+]ž Ë¡Î F “±÷:Hñ�´#ð²˜èFLqËäýEÛ;ÑùÃEétÕBîäé_cP#2�胀"JB§€z¢E�o¡Wb¶·¢j'mohèÀ8Œ¡®ãSGãûPð ã˜å0N Öùh| z#ÇÍÑšô'a c01Š\z�ÆÔºÓ*9_3ô1Jl5î�YÑœ6„½-ÏÓM®%æv9ÒúÆŸ3½ôVMGøŒÁú>¿.z¹ ¨q+ú•Æ4 Ûde“Òä<Û¥¼±Ðqv ý½jK$nw\òäb ÎBÅ™€Jl `0¦a�ùF/
|
||||||
|
£ˆ€ëP‹'ïu:ñ¤íì±,Œi0-ŒË‰3H‰LÄp¶Y› Ë!‰©À|Tøˆ�RqéÚ¤.5í«=).pž&Ã-ô–B.NŒÕROÆ4p0$ çñgœ[~“1
|
||||||
|
ØWtÂË;Û˜‡ãÄœÛd`£DÆ‚ø÷¨6CZÑ Ogp¢ÎrZ˜Ô“˜ûDK¦z>�ÕÌ©©a]p�Ñ1}ÒÕíëbÊÃש&EÐM!Ü3MÆ4èvƒP � ‚cÆa–¤‡\¦3H®u(Âý´0¥XòN#Ÿ™îIFÃŒ1¤Q2�ˆ7ºùǯJ 4±^�—c"�îl`wÔiØô\j4W ‘˜Ò²˜°Ëq'MI§!4`fðÇÁ �}½ëÌÁ»)*ýí´‡´ Œhàýã¾½1š‚›Ã¾MbôuW Ó£0•1ÖQÏÀèˆbb 8™E·F‡c¶Á�m™CåmþÎÆ‹,ËÇÌ(Ó€¼(ŽƒÍÌTNŽ£ éhÕ˜ õ³f
|
||||||
|
;¬tdaŒià¼1LjÑ0êP¾=Ž1
|
||||||
|
ü8¶Ä-+x” K�á"`jGqû#ÊzË[hk4i¾ðù˜6¨Ãë÷Ò’ Í@ ÌIgy³ŽëB{2x ‡1Æ4`0 ‡H—˜-IwZƘteÞ. t¨¤Ã��Óh®“f|TÐ�ˆ(‰1
|
||||||
|
ˆD®4Sœ+o½X2Í–"ÿÏ�WVwf§EA©¿Ã¼:šæAÍ»1 fuÀBá0*ÛUÊŸ°Nu&C&‰+ÃËØî-r’7 SXôË OŒÎG` wÐe⇧˜Fc"ž áZAÒÃe5©È8;vì$ÝlÄ&ÅÃ,ESkz“b‹/§S»eú¬»§I��œI‘ «
|
||||||
|
ÏÇ| ¬n›¤�PS Ë"¢^æ êf#Ô-ã8 ¶�«EZI¡iÆ“NÉ�wUqëç¡K¯pݹÝ';]Ñ(¾2»Ü®L®ÀNTŒICxrfª1)cÐ%o¯2~鮺A<S9p#fKdR*¨¼ÄxkÎá#™Å‰Û” ËÌ‘Ô¥Hy" hP £Î‹sFóš75±“�î²&±œ‘Ä4Þ.óŽÉ/Ø4]=Ž!�R²(Q!+ øšse oÇxs"O¯)9§¥(�ŽM¥_cÀ ‚Ô¢Øh$â0!³= p*4¾Õ0‹hŽG2lQ¨ˆÉsÔxéò„bqÏžá l4œ#¦,1W ;HUhxCy‡™Äö`°×:˼_ixウ ¥Ü©®l�VªÉgÆ”+f|Ò“ lPáqx�©Z��ÌQÑ6?å¹€Œo$RÀ�¾¼¨Æh*‰ƒIc�ÁRK(
|
||||||
|
ë)c…ðpÐï™+ŒÑŒ0 ™¾†m{°:!°Ý˜QKa=;É[¡0œC"àÆd¹ƒù!†
|
||||||
|
ÃY™Ì7&ÙaÅ€Å\ ¢˜:Cˉ{劗õJ�°]¹,Tw˜
|
||||||
|
{X|Æ3œÞÍ,Á1LA gÀj�Ãݺ_�ÇpL-.Ø`вXƒÁÊ`&¶sЩÄÆ>cŽCþ¾Ì«ïjBk7¦[â»�púè u0£ ›—)œ„’oxÜ#Ê�Ÿ
|
||||||
|
ŸŒ^²¤ ²��7¦“BÂ*R#L[à&ê$¡›ƒª¸0eAWé ÇôUì×쯤Vɤ �A³‡yÕ“ ×Ä:½û��5D±
|
||||||
|
á]¥JAƒÞ“¸)73£½ E§kå�y»P”û9£&+�y,lñ–à‹¨>n Ô0x¸*TôOf]T?]Wnk·†Ã¶$t
|
||||||
|
Œ zÀ2YÓùò{s6^û�UªÏËòèŒd(�Ÿã:7
|
||||||
|
BÞC\G"´o•›Hö(³ý-í#8<ôc0•gt„·ò7ü -,–³I¤_5…Jz�¼rf•CÖA<{·4Æ2˜ßS8N‰[òq c,«ƒ;@=èN×8Æ2¢ ”²æé¢éòà›'šÄ�Ùö'ËMšE¦M‡}cøžYú'�‡çôï1–1é¾Äºpáñ륟þ=Æ2Ñæºà…åãadëLÁÃÜX„�8™#c,£:8†æD aŒe‡ÖX,ÁŠ¸(. Æüµ“â‰EƒÉÝ'… ¡‘ ã8¦£ËT×wv*ð0' ×…n.F´s¾0fvúP-gX÷þìYÊK&ÔÊ®dçØ c,£1á’ã]è89Åsa,#8ë`tß ³Ö#÷óƺ›ÉººÓZEçö�ÖMã…õ:Üþ=D¿N#n5¨ÓjbW™§‘ ˜îá“äôEV ¿ü?°qå„9ù:Y�„[‘™Ì:G�
|
||||||
|
\ÓxþKœ°ÔPÌ- )Pô%P†T6ÅzÌj™ÚåhååÈŽq•"
|
||||||
|
ÂRi0¢qâ$gŽ£5¹sÈÐÆí}g]˜J“³òü@Æ`ÓÁôຨ$ˆ:”516¤ÄµE'ý�¥mƒÔøÊSU•€0,Óõ sÞAb¸reß�UvPû˜$`Œ�Z¥‰±ª�µÃ/ƒ
|
||||||
|
WBÞX È[¢/F̘û͵‡�Ôü¢nC8wì-ÅšE*ɾ£Â[)ÂÐ7ÖH²~«]ódLCiüx
|
||||||
|
‰¯�Q’ýàÌ1º±Ö‹Çàµ1Æ4ö5ÛG$&Ýàî<üöhƒÓ ÕÄ ®†^¼š´Á/G\¢¡™”�EVhƒ‡?ê–J‰IuÀM¾dN²3wœÕZ•ƒ9]ß•˜¥ =½ «A“A�A ¨§ 5›Ñ,•ètd Ãy‰‰Ö£š3 f~º Š-öªÂLôßž†’`Ï+$?Cœgg
|
||||||
|
½¤WÖ!F4°Ç %nLsÇŒ›£�„T|ÇB,ê@°ýz\ “›jòz+7-A¹ÒøŠ”:X(´ð¹ªÌAµ|2œÑ° PŽ™ªÀЃk¥§XžZÌwÐ]îÀ” 3"žÅœ‰F7¾Ä¡I“�¡{î3Ìç˜Z(¼,åFñŽ½q{EU² )\ ò<©²k�”Ï€Û„éH‰^ö›ýœ
|
||||||
|
áÔÃ+}ôµð"ê†ø%MùG„:«û ÖÂý�‹¾ú€bâl9 !é�SÅ tÙá6‚©Î¨=
|
||||||
|
:c—‚¨ÓƤ†râY[š�-p&yX1ÆÅ)ˆAŒðpó‰3½š_¢ž d+sÄþááµ`R×k$Ò±Ö@´ìÇ{zdMJŒÆëœ*SÏLpv4k–ã•�*LÇíòËr&ÑÐ(Š TA _(\¥Q�Ióå4ÄèÅps –ºàa¦óg1™™NP•ÄWÆ5ŠgÈàôÖoa; tÊÄ3t!m$ ™N2ÄÈ~zV
|
||||||
|
›7mŸÙœE÷Á‚(Ñ°�Y#Û8+Žˆ Áp¨bÓ£pC Ùo?3lÑi0i¿§LŸ†ÔªšUeÅøœg8¨ H#Á–c=,ãäÍ�%]¢4¯ô5.=$ðÏqf˜}œÌNt°8ÂG ž·±Š ¸¸¸½°‚Z'†KYq7�æW‚Ël % æ5Fçu ™#ßý…MÆûI+y(óŠ™Í~,®hÃb�‚¢s9Ò
|
||||||
|
`B_%„™Ã2ð©cpù<•ÂÁ�
|
||||||
|
!¹�"ÿÎjK%>±<*Ñ`ò;ÄlõcèŒ@èÖá2S¼]\ßSiAÚîƒµÚ ¤¢SÖƒ$ –!e'¢V•¢`sB’ëüTº
|
||||||
|
å Ô36.D¡Jõ£Ná)éä‚ò<2¨ H9°ØB�B7ù½y×Ì@èØ9‘§lb�c›•™†9ŠŒ‡2Ïé†ö(Æ(èé…¶Y‚R›£3„¦««-јå2å¹NŒ0X¢á†pF<´hæÍ:ÚN{|Ýé[¦R‘‚Zs2OUB§,V—
|
||||||
|
,›l †ÔÓV¹S¥=Íp Cñƒ‰·…vž~‹æ4„1ö‰FefO:
|
||||||
|
Lˆ/•bB‰yÝS^¶Jc:Ê챎3»A¥-Í.Kk$,î oBj�v»c?1ïOåâòJKbGv/Õ'¶¦°¡Ò�f d2!, Í !
|
||||||
|
iyªFuªD®f"„“ïj#YËŽ´ b\º6
|
||||||
|
k@µ³ëô –À
|
||||||
|
Ç<!VÜZ‘"Ô'ãȼKNêQ�Jž %
|
||||||
|
³’
|
||||||
|
!Š0(Èz®Ê˜ÔΖ5ê¤�ëØ)!è“tüiTfŠx÷A®*¯¬ª¶#dkJ�LÆöThM£ðåGõcÉf¡;åÐ ¼Æ+xUÝŸÆ*oúÖ;one0‚Åæ¼M˜`à ±÷Zå6‡’P‹FAŒ—4J Í,ƒÄ U^T†+5:Ë y™°`žsa°
|
||||||
|
Øb"1WA´Ôâ\Bpò(nû¹*ƒ<Ìo¨M£,ù¼A+Ã
|
||||||
|
}fAÃeúþ¬ C(Ú=ÑaR¨1
|
||||||
|
AËýð(H·ì<ÅJ³–‘O¼QB,Feª§ V¯Ev%!j^ÁÆ=c§áE$ëi•æ5[¦•áAŒ@$_AMÉöT\<ŠžÉìë³QP`ÿ³šJÞœ¨%D÷ZQ "P FÜ�Á…z5ˆ ˜x#@œïÄ»„*ë “¿2ˆ)•¥z“L�d„/fyýÉ"ì›kqqö‡ç7U3ƒÇò(LÙe „\#Æ»„$u‘-\¥ _¨˜f*K´…õ »º®@ÿF`¤:1–Þ*–"Ç Îh©–½³^+¹˜]nìl1No�CÛ%&(…> pz;±¹u ¾‰ÚmSäY+¥Z„pÀÁg�]/‹V™ÿx1÷ÃHõ¾`ù’%ag¼áì¬,ð¦¶ã[Ã
|
||||||
|
ÑýDÑ\Ê•æß8º)Éh¢�„Ä…š™‚Î
|
||||||
|
§«[8[ 0¹Gc
|
||||||
|
ƒ.ÓX 7‚ª¡ÁPƒùôBïÅár^‰±ð‰Òh.ì‰!c¨F}QtB@g§wM/sm>D§ÀAUÀMsªD…J ̜ɚœ3ì
|
||||||
|
iHòÎ {Rj!SØ=M†&–„>Xº½!*Ùº@c
|
||||||
|
ƒÖâF¿Ex*‰
|
||||||
|
èûX®&»¿ñZ±X©Eú-!,£n\AŒ50�!™S“Æl¹½Á)¢G±¸Iž‚¤…ÉÙÄÆ»¼P©YúpÒßœTø<ØuT;œþf¦2äâ~6Ø4u^ä ‘Õ«L‰C
|
||||||
|
M±a\Ó{˜%fÛmÁÄöá„Bãʨ%Ƈ>›ßÜÔƬã$ Z,ãDÆXåÃü
|
||||||
|
µâ:riÇ� Œ˜
|
||||||
|
H¹‰±¹ÌlÍ¿Çxƒt�Ã! Î4¯´ŠY$M�\+¾ÊôÏ1ÞP,©�)d™ë4† ·ŒMjâÖi 究;š
|
||||||
|
½¾q3Ñ2>²S¬„QC/Þ
|
||||||
|
S!8ìœ ãúd!¯ÑÎwÂuŒ›0^³üÖ8*l9Äb�µ"
|
||||||
|
©4Äè\O”êÂÆ ì'"Œ¦T ×ïáÜX¢´C�1¡ÆÔkžØq¹¤Àè[•ãNX‰ŽÄpf&›Ð «×Ððõ§ç )5©Ž;‹åwÂrÀ4^÷w²nª´À\+Ùc ¥œ]â Éæ ÷ØÄu€qv4�ƒÀLª@3¦Q)¼jÐ0•y"ŒqšAÕÛ“Áýð6âa Þ\'ˆÙÃyH $0Ýé#ö4…¸ÛV´””£Lu<¹i‘½7„s€½„ÓaŒnA)Å-1æà':@åa‹r…hö(úb³³#¤lf1»{–8Ø‚/fð1©Ì;¹Ù
|
||||||
|
ë9S`£këÇ0X$¸¤üs 9ÌÊö[Æ苲~ŽÛéNoF™5Ʊ�ƒ3;ˆáq´,Æ.ß1z§XBWb s‰- Ð8YáblÒÙáUÉ 9(½<[ ±uLRX�˜‹c»šížNŸ‘œÞómáA£šª‹]èõ�aƒî$I}bƒý fò8¬+«uF²ÊBU?h ªLsÄ¡1´¯òŒÃãx"»½ÙÄ v¡èbø;Í=9ð~ñ5Ó<CH3Öç¥fn3–6�Óã ;tB…ɋˤk�[:?s‹©ìq,�j�FÏð•�ëûb²>C[‚iÝÊN‰+ˆmf¢æ›Ø`Fæh —"+³2 ÔÒˆ+(1 âJ¼v:cˆ
|
||||||
|
^Œ18à 415¾�‹¸2zÄiF�b©(½j,†=œúˆQEv7ßćŠÎ„á×hìÆ‚hNÀ˜=õÆn(§Ò‹}ÁBrØèh,k£y×£ $ aþ» óLRŒcF¤å) S˜,Ýk ;³ŠW…1Ú@åõˆaåѹ†Ø̬ /1®¸ôŽƒ
|
||||||
|
LÌ™1 OSì�oE=1—3†±ÖØË,m¨¨ìÞ-Hׯc%x¢ž!lBŒ¸ ·ç�›F¹�IM±ðke>”]à G —ú·ã.
|
||||||
|
±«ªòÄXChX‚ú¼²ÆaD°¢7«�³ÚnbC÷ føÍ °ÏŽ‡0ÌÀþa%†´‡ñÙhSÓ©Ò†GÑ]™lÁ5Õjßt¶ˆ¾ÐybÑM«úxoþÉüÇ>cà ´‚r
|
||||||
|
�}¤›6šÕ0ÇR «–›,nB¸Y° ëéCiÕz6ÚÕ‡{ß^E= íj¨ô¥c3ˆîWJË껾Â�dÅ-k\²¸¤|˜�W“·reŒÛ¾o 64Iö’›WÏa %9Q»U§2Æ�¤¶ ïYÕ�t/HÓ7Æô,ד´J‡j¡*ã軳¯ž*™m «:M–ƒ-~/î×ã†[ƃŒÂ³«„›^cô™!Ä&RIvB8³ÍY ø_fI¹œ°,i¦©0b-)À~ª†Cµ�Á¾™*ÿFÓÙ©~Ñ\S|%ÆÑçÔ4è¡lÎv*)Ú=c�±I¯×¸É@M
|
||||||
|
Œ×Ž§�&Ê >W ©_9߆*C”ž
|
||||||
|
Cj!p
|
||||||
|
ëêàcÈ^®aêV©Ål :00?c²éK²1ê ™‚ƒpÆŠ1ì Y]Æ5êZ²àkÌñ? •Ws 5 ÓT” 0OêÍ�1ûc£Ú5tzÍ“ �‹%†‚ë´šlÃk‹uÆ ª<B†ÆÕH¦1•ºÑAT/c‹ýZR@Ь|'1Mñ‘I�ã~g†éP`�Ûá9j†%…æ£ÊŸø±ÊÀ™÷SÁq
|
||||||
|
ÖO%w\%Zgô¡3!ÆEâ¹0Æ Ú©f4Âh}Ø/Ô˜ÂN[g÷8fêCën�±È*Ó-!ìT=¡ŸNAÕÂÆÆ*é5,La_ÛF;¼/Ù¨Ì~|B±Óý
|
||||||
|
XÄçTŸTqìA¸ êt64A¦Q³í¦GÑ\Ο›…vĈqŒžÍîqt�&aÄ8¬–ª¨ˆ1a_¬1¦r…ÊGW·šFT½^V¬1™çæmY#ï]C³TGÕO6a©· †" :åÃðç9¬ž±˜âÄ+ªNcþp�]ƒµªkª°È�Æpt'EÕÕ„ùê>ãƒñ†¨ÏW�<HŒå '[*�‡ 7¾–e2÷‡éá=À?>÷m“1 ïFµ’=ªŒŽ€:CùÂXËâÌo Æû Än�>âS¬
|
||||||
|
¼òcÔh2&¯WòëQn’ß³í¢4ÔÄÁÙc6´Ÿã¬ÒÔÝâ‚Sw²"êĘª ây2.Á¬Ì¸}X „ãéßc`‚ùÕ=ÆMwaÆÈD•u(Œ£¶
|
||||||
|
&Ø´¾ä‰°„ÔÄN»;©¯HŒbž™;6tÚÝLcH5Æ%:O¥;ŠàÄå
|
||||||
|
ƒîÝ~¨TUsSaïý(ú!¡åÎØÆ »éÜÄTÌzÄåÃ<³ì|�N»ûPß5…£dýP I²êÂrèø�v÷-àè￧º!cNP‹í€y^ ÇN»Ûñ¿3žo^Ù:°"™r°¼ázöDŸp`ÝZ×8šLÃßÉ0ô»9¯qv˜ VÝM�íkØ£a$1æ‡çcÄØŽ.NÞ²HÖ ŒUÃr;÷D� ªÍq½thåZa9‰j° ]IŽ¸‚ØA±_§` twOƒ�ŠHå2Æ8gÏ@ô¸£=¶ Ó�ÏÀØäJU��±_ƒ§ÏX…ˆT ãp;7OŸÁŠÌ’ÞǸ.³ÊØ°
|
||||||
|
› SáqÆp…ï¦ë®”/HæÛÊGx=8s f`¬£3„0½?Ýýé:óÊ;"×ÑJ|SÒ˜pªîò�qWBbŒY°YïõRä5c1;[ò^óÈ. Ç5GW�N×"OŸ �ì2�½eÊ5Œ�âG ŒMGÙ9ÕØIÝCfdg¦T*6 ŒmG²§ÂÐ4·Ñc\¡‡»¥ÀXBD›ÓS©œ% ){oa¡ qZvÀîf-ɸD-õ�éX3Œ¯ûJ{`¼]s Œ¬j7ƪk+™XI•aÖ¸^YÓ]MÐé�¢
|
||||||
|
;®£Åt '´u{BÓWǹãtù®žs'Y£z`C]˜…Iãaö…¡÷ŽH„æÕo@þa³°ítª_Y¨ê-ÅŽÖÙP}o´, ys¬¢³ßÕ9La0븰~õwf� ÏÂÈà©9bœK¼<Ž‘ X…½Ä8z€Žzal4[{(wxüPudYŽSÔ1!ÆÍ <ô�»t)w…¶æ'£�X”�Uc0|¯—PÌPnŒËŒ·ÈÐ Ý•=¯c TDè þ=šX…¡XH¸÷X'Þ`gïqc¬S>}—&Ý&Gê„©'MÌ“I©lO1cœK�/ŒMvªæ8›v9—Ü@¿F©&ƒ«“ïX rDH¢›~`.™ë —yÕtf ã¤MM —Ø«Üuƽ0žW 0¶¥‰½+Û�ñŽÇp“›ö£u’[�qÌ^qYc¯ŒhˆU@ˆ��Z1«�1þžå%$Æf³³Æ�UÑój’•:)¡ÉÁKéTe‰c<Ã=.A3¨ŽÕr©`ŽaxÁ‘†Öi'ç| ì¨àmDóû€6ÙÒ…u5ÒÆpFgª·Xºš ÃuÏ„³7™ÝÇp†Ú¡x ëO#„Ði~ ÒÿG`L+={`tPOc|Tch™i»3] ïß…ÿ�ö·»{\ãØÍHÖ+1}¨RÝ[3Óuø¨7†3ð«5Ç8è«’½ÄhJ53R£wß[ŒÕ¤îfã-�£›½0†3Ne&ÆÂ$í‡ÆhÎz=®a•yÏ‚Ø7%)n P§"}a*Š)óÆüP©½�ÙõêÉ÷ b„=<ŽÁ õ5‰qÌ}olÈ,È×8ØN%VŒÁ 5õ9c.…ïßÎ`ƙջИó®›ÜX¥]ãÜž\ûH3Ãã¸ÆÕ¤ö˜ÄÐpƒð€Îäþz„wV*¶ët1Y-"»ºitç/ô^îîF8ú5“y�åÑA •D„F�˜ñà%*\ñKLÛcSëå×Y�eÑ{-rfH&�5•Ú/Gža fàÕžq$±-
|
||||||
|
ÃÃÂÌh§úõÆ Oö3c-îîäXâô\¥þƘÊ3Ïyí¡G‹¯>f)*·Ž]ÙèœÆg5²¹Æ¹?0èƒ}4®�Îìq'ƒ¬ùzœ,ºŽ_uæ]1õL1ŽŠ¸,˜PV«S½�a38/²+®4u¯ˆÓ
|
||||||
|
¹îea& “ïÏëde*°Vr˜¼È†8yÎKÈèŒy]ó 1L{ÅþßΤaõŸ ×0¼¢Ø{,šRŸëhÁnW{;c8tx ˆÞgª¶ŽìC/¿GÃáØaô5�žæd0£<D4ÕôáTLBÔ^Ô•6d挔P ˜íEÍ0 hÂGHX'5ðÃÑó‚/þ©æ†
|
||||||
|
JEÛøk˜õc,Ã�7
|
||||||
|
¥«„ ÚÛ»$…¦}Y°‹yÉ�Q=nñØ ^ÉwNˆ¡Œh a¬ÒPÓ2³bÏÙ6qÇ°%žó|ˆÍð7FN›˜
|
||||||
|
cR@b ^æáæfã`,£Kð^Wa�$EhuÔ”ÔÎ$0l"W ‹†ù±bt{8·
|
||||||
|
ßÍ:¦¢lã¸z£]ø8Ô9«Ì ìÏË•nl¬3©nc0ƒ—Öõ‚Ú•ÁŠ«†[X/: SIC�1–q¦t
|
||||||
|
VÆ€Fô 5Æ4÷ê©0 ÁDÃPîsç]Ü?×ä Gyè#W b¢
|
||||||
|
8.§rwc0À“óz+*û%F œ-Ð�q�Ë=:£©HGB`éR^E ~uá µé
|
||||||
|
–‘T× n×Ö¶\±ÉÏq×(‰0 ¼¸cx �nYô ±p-¦ Ï} —bTêÕµr\«[®®e¿ŠíÇTc
|
||||||
|
Ñ`�[{æ»0f½Üv®1CË9MC¶p�]Øá€ýÆ5X¨•j, ÉêÕX`5:ðlTc�•ìV ÕX`ç‘ꪱÀR°k¬TcÄxÑm§„%É,;+ÕX`£[5Þ¨Æc)Æí‰i, ¨uf¿Z™Æ«¬‡½=1�&~…'¢±€x®ü• ÑX`i¸¸c# Œ%XóöD4fŒš£i�Î…h,°q˜9a# ìjöáxZ�†k¹×ÛÑX`%*Ð7¢±ÀÎÃaÝ�h,0I¼ÛÑX`2nODcÆäÌ0yÙB4X¯ç=xÆj‘e¿�V¢ÍF4^Ÿ’•h,0ïžxÆ:DëôÄ3f ºƒ³6ž±À°ðåâ {ðŒV…â Œ[Ðü] ÏX`9;U|ã
|
||||||
|
Jp‰-<cÆê ¦†�g,°®ÐÞÏX`.Êzâ ¬°ÎäöÄ3MÆž±ÀŽh^º�ÃÖq£Ñ�h,0*jÆ¢±ÀêE¶¸�v2Nx{" ,Ų�; ´{+ÑX`Ì/=oODc�Áìõ~Y‰Æ»bsÑX`
|
||||||
|
6ÝžˆÆŒ©#¸ÙĦ±Àú;™Øƒh, é%·'¢±Àò<‚Óm! Ìm~žˆÆŒ¥Ñ¼*+ÏX@-Úÿn<c�¹ ÞÏX`®Éxâ3Fê¯ÊÊ3I»ü� ÑX`rŠßžˆÆË)¨àV¢1ç]ÌÙÓi°…h,0g}>�¦^Ò·'¢±ÀXh‘oODcÆÜâöD4ØU º�v5S݈Æsc‰�f̈›,ÜžhÆs%ÃÍX`yT?ÝÊ3ØÁVé·'ž1cŒ`™3qå ¬(Aã‰h,°—ÍÊ3fˆÙ™ñà ÏX`UšïÏX`):m<cÆÔ�Ó�¾ðŒ¦Ò€ÛÏX`”rfÛZxÆŒQ°™±nå ¬*åè‰g,°¬ä¡'¢1cÊOð¸…h,0ú| ,å þZxÆ 1
|
||||||
|
Î;bå ¬$ÍWž±€®nÞÏX¤.=øQßyÆae³lá–æÕ8pã ¬D¬�g,0Ö ™:ká36Ø‘ù™g, ÕôÝžˆÆŒuy=�ôà…]‰Æ;.ê²wž1#d6Ô«YiÆÂ� _øN3f¨úD<ÑŒ–Ù à¶ÓŒ‚Ét—Ø;ÍX@ÆÛ•fÌØ9"�x¥ Hwáí‰fÌØ#x°ÒŒ„‹Ïd +͘1Çñ…½ÓŒDLó~-4cÆu4
|
||||||
|
ìA3Ø™l±n<cÂf"»QüÊ3Nθ¸Ä <cƆòÔ{ðŒ–ªåìÊ3f¨×‹Ædå L<À�=xÆŒ=²Ï˜1†‹Wl! ,SJö 3†{)›f% ìªç؈ƌáyì ۈƌá†qg¹�h,°tq‚-<c†ì ¶ðŒ#¿³iGVž±ÀØ!Y˲òŒëῶðŒÜž=°ÏX`WvÕÆ3f¬»‰èÛÎ3fŒM˜ÍX·òŒc¹ÒGž±ÀpYœ&âZxÆŒÉwiB…gÌ-³Æ¬<cÆØÉ´ +ÏX`)šÏ˜±$J´À <cÆY0ÃJ4&Œ=ϬslDcÆX–âí¹�–¢ÙåF4fŒäzE+Ϙ¡¦>³Âž1c‘mõ¶óŒs=œ°…gÌØÉŒmó~-<cÆ®Ò �fÌPößhÆŒ‘a\Tbš1aÔËÎú�f,0ØåÁßµðŒ‡ ð͘¡~„b£3Öõ^xÛiÆŒ•ÙG<ÞB3fŒŽ�ø½…fÌXv[ö·�f̘\› hÆŒ‰íÆ߹Ќ c[EWWo4cÆÆÅi¹ÑŒ£zeY¶ÒŒ£$«•ØƒfÌX°½í4cÆΑ]ºñŒcÜú¸¸Ä <cÆÒúÝÆ3fìQQ´1� cJÅÑ>0�#ù‘yWV¦1cøõ ØÂ4f Xp¤¬LcÆ ˜ÆŒá¥XUؘƌ 3™àycq5ã@¯LcÆx¡zW¦1c̳ñ=µ2�SìÈü^ Ó˜1éÁeºR�Ë)dàÊ4fèQÓ¼1� ;iöÍLcÆ ¹mßF5fŒ•ÇæX[˜Æ ×$¿LcF° ÁE³2�Ã�íÁBöÎ4&(;-žÐÊ4f câŽZ™ÆŒÕ ðÓ˜1HewÝ٘ƌå‹øká3òH�ÙxÆ„A“»øWž1c�ÍE&ö 3Fh^$d¢1c,°`Y‰ÆŒQñòM³� ƒ-íà DcÆšËsÞv¢1cå²ìW¢1CX''mlDcÆXr�Ÿ‰Æ5²L9<´�sŠ°w¢1Cî»ÐƒhÌXºò%W¢1A
|
||||||
|
N›�j%3†«8ëJ4f gÌÑ��hÌ3,bÜB4&,9…-Dcƺb¼íDcƬxۉƌ1Fì=¶� b š`*]‰ÆŒ±‹ŽUû…hÌPq3Š·�hÌ3u|~¢1Aâ) ˆÆŒ);ïb»ˆÆ a%»Õ••hÌ‹>ðŒ Â)ˆ²®�gÌsßÇÅ%öà3¦²ó3°Ï˜°Ó¦»°…gÌ$ÑqQ‰=hÆ E7η�fLØ¢}ÛiÆŒ± ¡| 3–¯Æê͘0\™8= hÆŒU ôB¯4cÆÒ•+±ÑŒ ƒph!ûVš1cïüÁ+͘1× [hƈ±dW†·�fÌØyå'l4cÂÆ#”´ÑŒÃ}ž ؃gÌ´e'ÇlDc ã]âh%3&Sj-DcÂ>ðŒª)Ûs´�;˜eÆ°…iLX}X«Ó˜±|ñ›nLcÂ
|
||||||
|
©¶Íj²2�{t¨Ü˜Æ„�Êz7CÙÂ4fì�Khe– ¿¶0�; ýíW¦1a,J:M©³2�“çíb{0� Ã9õݵ�Â<íÜۈƈ±Çrr+‡�h̘³”6–1Ã�?a ˘1æDš0deF… ©V’1CL\
|
||||||
|
R³…fLX+WGÈ�fL;8š¼ÑŒ#EkèÁ2&ˆIÿ� ;Ù±`3æXJ`¦1aYF„oa3v\†øÆ4&,=’36¦1aÌ‘tõÇÆ4FŒ�©›K6ª1cLÞ-�؃jLØ`s¢˜ËB5&¬«6äÕ˜16ðóÎ\©Æ„5‘N¤À \cÂÈ”,+Ù˜°Ò]üû¶“� Ã{h®XÞÈÆŒ1ûû¸ˆÈ dc Ü#‡ic–"ûímgƬ‘3h¼¶1bì }Æ7¶1aŒ•Ò¹°�;äìÁ6&¬“ðÃÜ-çB7&ŒÒ×Þ7º1a•ŽuóU¬tc ì[´¬tc°Âú’·�nLóSGP‘-tc ÅÌ‘·’� a'ßæEY¸Æ’V�z¨Y]V®1aál
|
||||||
|
ì"ôž‘¼‘� c¾‡iÉW²1Alë?.²×˜ “MÔýÔ+ט°Ìíçûiå–²S‡W¦1 ñÜÊ4FŒK˜cE¦1A4ÁMnLcÂØ~÷ð©[™Æ„±»ß ò²…iL…pPTc ôàzY©Æ„áwÔÞ¸ÆÔ«aVµaÛ©ÆÉy郼P� b�� ÷V¦17�¥ôÌ4&ˆ $gÿÀ4&,³U³OÀÂ4æV®t´0�©EK¯Ü
|
||||||
|
weÔe ¥€ LcîlÑ™ b‚²…iLYµl®lLcÂpa "oLcî¤Á^†ùÕ˜:p°×úsY¨Æ„¹‹™™ºª1wî�Q:{P� ã-ïëg¥ÄL0ã¬TcJð™b`ú@5æ䟕ÀB5æÖ#ì ê7¾R� £¿Å¨wª1w39… èA5¦.(3Ê_Þ6ª1AW2ÑÛÎ5æÆ*
|
||||||
|
d~às?–ÐF5æ6.ïRt¥‹Ö08¹~;+Õ˜[Ѻeõ�+Õ˜»Q ¤LÑ4Wª1wÅ©ÕÕßÓ˜šé¼7ÄؘƄUþÌrµ2�%÷unûÛN5æîWéê´³Q�ECï¹>Õ˜[1åÓbc¡”Ø’-¥€ \cn>DçtûÀ5&Œ!·Óïu勆F�gãטº щí&ÎÙ˜;$=<odcnàÅÌ%Ëî•lLؽ(TЕlLYpì8ä€ãF6&쬋d%s¨GòéÆ6¦ÎQ�ò§`s7ªƒþ8�[ØÆÜÝ ïõð=¸²� Ã2 ,laÆ.ØA>¹²�kÐhä{zÛÙÆ„9q=ö`Æ|²Œi Û±Z…0û×Â6&Œ%DÁ¸µ��*W'Ê·�lLu4ŸÙ•kŒ ){ÙÂ4&„ªï laSæ#ËöÜ�dcF5.ô³•iL“µãÆ^™ÆˆÑ{9ƒúkaÄþ¨¡Ý¬LcÄöO�t¥cò&ì‰<a Õ˜0¦ Œù�jÌ-øI›ã®1a'f.Z¹Æˆa^5È57®1ajÃá¾þ+Ù±ÆþÜÑ4~%F×÷i©•lLÉ©Úì Éƈ•°W{�� s•I¼²1bg
|
||||||
|
[Èƈefh\ó\ÈÆ„±˜)hV²1b)ö´°…lŒ¶ë²�� #E}ð6dcLÚÅ›=$KÞv²1%ôr§^äY ÙXR„ñ!6²1a‰ (7ö_Éƈ±ÛC±Gz##Fý$»šacgC�>
|
||||||
|
oÛ¡S�÷OÿÜB7&,EñðÛF7F(S¼:H´Ñ�)Ë™
|
||||||
|
¤ì—ØèƈQ4ŠÜêm§cæôdÃt÷6ÙèƈqߨOÈÛN7FLÙÑÇü@7–T7v))°Ý1ö=í07ØB7¦,î6$O„-tcÄÎæ›OØ;Ý!nÓâøæF7FŒûˆÝ_…½Ó�‰7ã½Øv£c®ù¤Þë. +Ý!†KÏ# SŠúQ ‹y[éƈTÖmùü¬lc„È/“ÎÓ_Øƈ±sJ;ü} Ù!Ö¤Ð#Hl%Sý¼*6®1&ßc:«émçKzLæå™÷ì\ÈƈñZd·¤À.²1‘Œ¸ ŸÉÁ²1bL)f{da¶1"Ð ˜ŸRy°�%½ è±Á ´Ò�±÷ L¿Ã
|
||||||
|
ÂN7–ÕÇ„t?÷B7Ærnq!ïlcòË ã°OygS}Æ®Mc³±�eUN]ÙJ;ÛXVq8ôB¯äF7F c¹Õ˵Ñ�±™Ï ,ž�½Ó�±iþR�·Ò�‘l01Oþ¼(ÅÞéƲºc¹µÐÛݘh´oLæµÒ�eµ=sýåÛN7–Õ™iÏþÊ•nŒ§�vÉâJ7–åYP›#A+áX¦ù:Y±fê°…pŒŽ:Ö²Çs¯„c$S€€Â(S°„c™>
|
||||||
|
rÿ™äe# ;{ýXÞo„cän`ùëèÆVÂ1`•õ.¦ÿØ ÇÄß>‚'w' #SDR=@ ì�pŒ¼�¸Øf§£Ó“[ó�±Ã&t-vì�oŒMÎd6ÓÚÊ7FùÎ k'í|cì,ÂãxõÙÊ7FÇ-ýí±&+ݘš„²Q¡/¬�nŒIY ÒÅ
|
||||||
|
²Ñ�‘ªz²»ûìlc‰õ£•éMf"[ÙÆ’bùŒTyÜÊ6Æ~+¤ñÉób"{gKÚ¤Ê඲�±‡ VqzçÎ6F9S¯Ms²³�±/ £,Ó*ÌÆ6–t– Lš6gcc¡[š—›`g“å)Ô·�m,)s5ÜÜÙÆ’šÒ±Ù¿™ÈV¶±¤£\é“ ì�m,)€ :fñZÙƨ
|
||||||
|
épâÎ6–XÃÌ^C~¾�m,© /îŠþ‘mŒvu=Î�lcÀ دþE;Û˜›úŽþÎDöÎ6FÝ‘í ‚pcc
|
||||||
|
&~ý0_ÂÎ6Æ:KvºL¾_7¶15&–;ý#Ûõ_z\²½³�)pu@tõ‹‰ì�mŒÙ§\Ü�\co¼ÓyÂ;Óô÷Âb¬¸J6¦1•³b¾¡„mLc‡JvÖèÎ4v(bÓÙ@ØJ5ƪ[ÆòÅBöÎ4F‚Pö!-~´�iŒiÓ%V^æ•iŒNö…<å>ÙˆÆh˜±ª=9ÏÛ„±O¨…ØÊ2ÆÒf: ‚ºu%c`·:¯/èA2ÆþâìlR¬<¯ cSÙHÜ„ (Ʀh$ToЃal*á€ù[æàZÆøÏØ0¤›³råÃïv–©Mo«•^Lß™o™>°‹1_ù�¾çVr1ÖáS' &Á•\Œ/�~#ø¾n±¡ÂásôàûZ¸ÅØ^Ÿù v+nÜbØDÌ ÁL¸r‹
|
||||||
|
1MÍ ½+·{ŠõÛ g µ˜Z+ ìsAj1ud q£ú•[Œ}Y$ÒÛn1úÊ.gæÛÎ-ÆÎŒ&ö6?×Â-ÖIÞAv¸áÿÊ-ÖÕò”if ¸Å8ŠÒÖDõ· �ϼ<A ·g˜iê=Ä ÜblPÁ?»ûÔÆ-F? 7‰Ó
|
||||||
|
7r1r6@¡ÍÁšµ’‹qyσ}ô cÌÉ5j%£û‡6€{öoìbêwßâ~ÝÙÅä[ ì©Ð;½˜|þ
|
||||||
|
6 ZéÅ&œl=lö®•^LÙ’´Z@ïôbnCzšŠ|§c»Gúkâ·Vz1Q6ö1ƒl¥+¬ÏÃqö�^Œ†¹'³œüb4EøXf"ØøÅhÜàNu©úÎ/FkÉil‚V~±ªîò¦½=ñ‹5z�`›oÖÊ/&ÈÅ(oOüb4NIdj†�_¬)é²ÌÏùÅÞo‡�`¬‹;äp#¦�`Œçk±
|
||||||
|
½=Œu–õµî¶ž;ÁXg?>5»´ŒÉû@gÎÅ"öN0Æ{÷„{ÍîcƒM7Z
|
||||||
|
ÞŸ•`Œ ÛNÞ^Œ•`l¨ñÕƒ7m!¬i«#^äF06˜-£ê>A+ÁÝWíŠxîc“ÕèW‡À�`Œ:ýÙ†Ý4;Áؼ«‘� Õ6~±©Ö}ت) w‚1µxOܼ‚‚1Ñ…@7–B+Áo?¶R±#i#Se¹?ƒ±l!c›hóÃÂF0&ú6H<=j!ã=I7[Ši,c¬ôbçdSVlclÛB²ô`[[ ÆÈg‚»‡+ ÁsÓHÏ ÔN+Áoò ëï ¶…`L}X3ÝW- Áouþ£âM³Œ±Þ,Ñ{~ cóÊ;ëPÁ˜XUØ¢ÂË»Œ1AN
|
||||||
|
D7÷ØB06Oõ c; A Á»Á¾ÓðübôéÓ{3‚má›j‘‚·r^$b~1v°a†Î
|
||||||
|
¬…_lª©ãîZŒ•_ŒIzÔÛ�"³ñ‹‘°�üÝ=ø»‚1õ¤e Ò ] ¿ã
|
||||||
|
™e ¦û\ùÅX{Çx×a.�•_ŒmtyŒƒl!» =µf \ ƧàÂWéJ0Æb>ª–ÇzŒ±/Oe¢}<ñB0FE+“8Ø$A+Á “Ô?ÁöŒ©8ðà*š k!³2ÌfAZ•`ŒMxÉ;|ƒ®cTêØ1Ç
|
||||||
|
7†1ÆXÈbë´Œ1û‘¶nO°Œ±DÑE�3 Á Â=h{v%c ¢£¨ƒ‘�…`Ì-Šº-†�`LHe>– j!Ó÷I«1´Œqì¼_‚Qm!ãÔaT7^ÞƸLºXÄ‚1®-¯™þ�`lÒ?Á~j–+Á˜´lyø±V‚1î¦Uoo;ÁØÙ+“ßyð‹±mT¥àm±ò‹Ñ¤a=_(üb3¤ÌlçE"öàSì�
|
||||||
|
Bâ©V~1mGè…ý#¿»=Lµ�_LÍq˜½ó‹©+t ›w㣦†³Ùâmüb“ÁNóÕôN/FË›o¦ï»�^Œ·}c:•eáF/FY.¶ÁÞžèņÎîÕ.q£c©(U/ÿÞÂ/Æú
|
||||||
|
š?ÌW¶ð‹±bƒŠ|0]®ôb*ô ñ�Ë£Vz1Ve´K•yÛéÅX]¡bûiVz1
|
||||||
|
³cØ߸Ћq"Ê”õ%°°‹q 5ˆ¬Þvr1bšÒ¶Ø…\Œm‚&“ªäb,9ß¹ Vr1Ξ-N£ÓÓF.Fì Åüi›h%㣩_B\Š+¹§‚:Rh¾+¹˜*fX-?MPº’‹åèx!/ñÛÎ.¦—Cϼs/7v1~'V¼õ ß[ÙÅ8—¥ÙÛÆ.Æg8)z.laã’• (ìÁ.æ7®T Øň�Œþ´j6¬…]Lë9É´.}vcÓ+â%xšhesM�m€·�\Œ�ÇÖ’=°¹—…V*ã®Âr1.çd²¾;ÅläbZê¥
|
||||||
|
ýJ.fìp›ä·�\LßéŽ&%[ÈÅ´Û'ÓèLX²’‹ñ2 ô£ 㣠™þ« k[Èňá(³BÉ\Z¹×²ºgwàA.æ=mãXØ;¹˜^ óÌŽ +[ÈÅ„©—¸Ùkv1}ãe¿íìbšIkÎA{ÛÙÅt‚”Ñ:ü• ½1èÊ
|
||||||
|
ìA/¦ÓEG¾‹b7z1I7v÷3ýÑÊ.¦ƒ÷h¯°‘‹éÅ%¦—5�-äbñÂ)Ö½( ¹˜ÕŒVlo;¹˜ÇEZèÛN.¦ßÿT#ø·�]Lg¹éÁÛÎ.FŒu!6v1m"÷�1ÃÖÂ.æû ^µ´»˜ÎÖtK`v1¾=h7ìR+h!ó›ÎËyÛÉňÍ-xßvr1‰[ÈÅaåj#Ó Ëåq–Wr1�cpþðþ[ÉÅ$ŽHè"ð�\Le�¬çi ÈÅ$áŽG-ÖF.¦½Ië:� ÈÅ´bé‘`°Ñ‹y¥³$º°…^,‹ÅGí¶MK¶Ð‹ñͲ‘F:nôb l{RMÁ³Ò‹Ãý|eÁmôbÜed´ïN®ÝèÅ´;+{à™�kåÓ®f¯ �_LÝŽlübúÎCÌ Æt³]—&¿ZÆt| \,b†1�ÈŽ³_äc‚±ì†Íóp(g#Óröçô4W‚1Ý"¿43ÖJ0¦×Ç4$w´ÜÆ$’XµZL µ2ŒIÏ™±Ußv’1 "º¼#V–1íÀ‹[çmg#Æ;ˆœ,–1 0šÙÖÆV–1_Í�U3þÊ…eLgÈN}c½³ŒY R5ðE@ö`ó‘%ÉàùÌ2&Ñ?Ü]UÐÂ2¦“7ì ìb#ÔT2ÐÂ2&= _fÊÆ2¦S =ÿÒV–1«š íx�W–1½:üy˜´p%‹û€Ž1Ý3 7úˆ¬cRªØ Å�¢6Š1]Aô£¹=ðF1¦={æµâÊ1¦ $,ÄVŽ1©½ùýõ¬ c:æ�¦-èÎVŽ1_BL(4ÃàÊ1¦£Ì"·¹Ø8ƈ’ßØw¿qŒ c‡D3¬o c w:–÷¶“Œ…X ënã³4Š¬—·�dL’ƒ²ÇM¡7’1+\VÕ„-$cÙ¡ywHyÛIƤ
|
||||||
|
Œê$¡·�dL:#V(Õ`ÍZHÆ,¤y› Ï$cÖ²ß4[ÐJ2&�Ý2»‰ÙJ2¦BxÖFXqßHƤùÁˆËv(,,cRÛÝ*Ô¬_ ˘Ž�5 c ˘ŽO
|
||||||
|
v ·�e,«Á]U
|
||||||
|
”°…eÌbŠ-õ‚-laÓQæ®Íó˘$@a®õ€•eL"l¨É¢Ï–1É°oÎmÊ6–1aîÐù�fL‚ʉ-fÔZˆÆ„1«æF…hLØAæ ïè•hL¿W˜1u˜¼l!³âÛ¶�éùȹkŽ®…fL‚}ÐŽô]±ÒŒi5S¤¯¼í4cz ê)gãiåÓÛÃÎ #v㳡
|
||||||
|
Å)®Þ•gŒUõjÇÊ3FŒì¸3ˆ'Wž1 Ô:¬g¼í<cÚ·Çû-³òŒi¿ã_jÒoϘ¥-ƒûžÊÊ3&Ý–&yò]¸ðŒ©Íóh\-»�YÏ«Út%ó1¿8 6¢1©[WÙÌÛN4fÍÔ¦Ù{�Ià8GÀ“\ˆÆ,¨ &C¯Dc’ßL<b– ј ìÁ”.ìA5&¡™µ˜"X©Æ,kS;ìÐÛ¨Æ,£¯Þ+Ó˜ Ö”;,·1� ;ù¥§Ç-LcÆ 6¦1aEz̈¶0�ù*a ¤`D[˜Æô9„ùÛÎ4&l\ R7¢1J‹Fo;јV“¹ENiۈƬ,ËOÞ{�éí½{ 7¢1¿õÌþ(æÇZ˜Æ´[(4ƒ4gaó
|
||||||
|
tYüјt†¨Û3Ϙ̿Äî;Áj¶�i¯�È,zÛ˜Æâˆ�íÕ߸0�édù‚3eØB5¦{‹4I.ûبÆälSŠÕGª1_j¬;örTcʇÊNR�=¨Æ(8Âä÷#,TcºòÞ[�oTcTÔ¨œ{¶Q�Y1?|©mDcrž@„
|
||||||
|
'ÿlDcRÙÓxp÷¬DcêA£2*“[¬Dcv3¶CÂJ4véú£$çÂ4&cÚv|
|
||||||
|
ìÁ4¦+�íÓ\o¾1�é*qƒ
|
||||||
|
çÂ4¦ËwZrö`ÓåApa Ó˜.µ¢d
|
||||||
|
³‰-Lc¾˜ñgúÀ4拲eu«{ۙƄqK8 °2� ÂA
|
||||||
|
}kcó½Ìª® êzg3ÄŸ³4:¦1«¬ƒüÀ4®jïå€ LcÂx)çÀÞ™Æì3ë
|
||||||
|
Ž•iLXV ο¶0�Ù®7åÓ˜Þ·é5Ë…iÌï5·‹\eeæî
|
||||||
|
˜Æ´W9ý� Ó˜öß%µÞv¦1íÛÁ®^>•+Ó˜T ÜÆÙ%Ó˜ÎBa§ŠãӘγšœBµ1�éì]Ålo;Ó˜T-›‰ke³�½·ßv¦1û±¦TºÀ Lc’ WGù·�iLªLV t3”-Lc¡‚CˆšÑre“$ÃÍ5L„¹�ɦ|o#¼�IpÎ<•Þõ¶�9l ?EìA4&a 3›C Dcr+±j9ؽž1Y% …ÊÂ3¦K¡±[’—kå“�S"Uïmã³JÅÞɦœZyÆt=eVªš§tᓺÅÆ&l<cºïÒÁõ—؃gLºXº8«7š1»sÙê4=ÙB3æxÈU;°±Œé¾.—b¾‘ŒÉÈfêÉé›w%“
|
||||||
|
0Ó>�Œ…÷ø,A?´’ŒIp?zÓi-$c¶lÛpÝõÆ1&ˆž Ón c¶éÙñòøÀ1f�p¶(¦ÝHƬ4¹Õ•°…eLžÓEɘ D¾&S¶$cÒßÆèÚHÆ„Ew€·�cÌ*aªQ*¼qŒ ³D/håÆþnÁ[´rŒ …°…cÌÙ¢ý ǘ¡vho cþ¹Çæ[(Æ<ɤôA ÅXèÉÎ~Û(Ƭ^¿³ßcV…ULdB³wŠ1+ìì¨í[de³–\¦#\ɘm 6ƱðZIÆ ¸°à¶�Œy±¯™±•eÌÚ5ݲû6–1Û*qàßv–1kÞÑñåmgÓ±}ö`³VÞzøÙ6–1 ;µ|ùÀ2f—²È�>°ŒÙ—Vrðjl,cÂƃëac³äx¿´V–1 Ì~`s@« ÔX+͘„˜;>~ ³ìs{³À 4c ÄÖÛiÆ$O™?U?ÐŒI çzõäÜhÆ„YÙ7öN3&ÁN Å05ÞJ3æt‚Žï•eL·nÁi×êÆ2¦Û‡¥ ‚¯,c‚ú|hv+˘î3¶Ç ñØï,c6dØÞKÞÓ�eL7d:¯–˘°÷´��eLwn?®xèÆ2¦»º`”Id6–1ÝñY=oÍßµ°ŒÉ8:è®êãöÄ2&Œ$ní½±ŒI§hb65¶ÐŒI±—z£“úÕ}TC Ø ˆ½ßž(Æd‡5WÁßž(Ƥ-ÑñÒ²Ç-cÒ²ÒãþÜ(Æ„]yrOcÒÜ ‚-d£“m—#¯ý‰bL˜šekƒmcÒ›:Ðœ·'Š1i˜ìF j+Ř°©N2ÏcRZy#§¶°Œ9”Í|�jÖ²w–1Ç,H}¤e ˘täúhí1BmNŒÊ˜äŸ¢K·ÿñÇqûO<z7ýÍŠ‚ľþIoåþñ¯ø¹ó [m2&�ýë¿â÷ÉLö÷ç@�Rñë$r¯ƒIíEgñ/ÅD8ä/ü :Üþú·Ûû·¬�øw·ÿ~ûëþã?üU�ú²Ÿbñ3¶ûÇŸ ¿è'UÞM×éó/ ¿ê!e¿òþýø“ÿò‹~’ý‡º|X×ù‹~’´C�]>þä¿ÿE?Ɉ�ª>.ìö“:,ý× –�A|�ŸÌ!}2ºîp1½vÌj$‡.æᶀùóIü<âí`–é•Gy!–[yçBÌO\ˆ·ÿ¢²Üƒõs'ô/vùblúáÿ¦Ä{|£ZÃn�åé?-à˜,ÊÖ`¼;±€û“ŸúUîO~ò ¸?ùÅ_%àþä'•€û“ŸüUîO~òÛÜŸÌáÛÜŸ�¢?pù5î~MÆ±î– SGžeÜ-2î+¯-‹1‡Qhµ™ùš÷O¾®/ü6ãE™Ùžß} ý9//“3&»m0?p¿Œè"={4@ù;—Ñ?ó¢¾t½øEýÙ¹úû/ê— «¯¼«íPíïêáú'6Õ‡™ ˜Y¢…~€˜Ùe‰ým±ÎÞ-¶ËŠ£�ÆÙü×?þÔŽ,B
|
||||||
|
endstream
|
||||||
|
endobj
|
||||||
|
68 0 obj
|
||||||
|
25389
|
||||||
|
endobj
|
||||||
|
69 0 obj
|
||||||
|
[67 0 R]
|
||||||
|
endobj
|
||||||
|
70 0 obj
|
||||||
|
<<
|
||||||
|
/Resources 71 0 R
|
||||||
|
/Type /Page
|
||||||
|
/MediaBox [0 0 307 179]
|
||||||
|
/CropBox [0 0 307 179]
|
||||||
|
/BleedBox [0 0 307 179]
|
||||||
|
/TrimBox [0 0 307 179]
|
||||||
|
/Parent 72 0 R
|
||||||
|
/Contents 69 0 R
|
||||||
|
>>
|
||||||
|
endobj
|
||||||
|
73 0 obj
|
||||||
|
<<
|
||||||
|
/Type /FontDescriptor
|
||||||
|
/FontName /EAAAAA+mwb_cmsy10
|
||||||
|
/FontBBox [11 -215 942 727]
|
||||||
|
/Flags 33
|
||||||
|
/CapHeight 0
|
||||||
|
/Ascent 727
|
||||||
|
/Descent -215
|
||||||
|
/ItalicAngle 0
|
||||||
|
/StemV 0
|
||||||
|
/MissingWidth 500
|
||||||
|
/FontFile2 74 0 R
|
||||||
|
/CIDSet 75 0 R
|
||||||
|
>>
|
||||||
|
endobj
|
||||||
|
74 0 obj
|
||||||
|
<<
|
||||||
|
/Length1 1992
|
||||||
|
/Length 76 0 R
|
||||||
|
/Filter /FlateDecode
|
||||||
|
>>
|
||||||
|
stream
|
||||||
|
xœ�Umle ÿ=÷ÒŽ{éÛ`3^wLØz×s€Ã%ë®[ƒŒ"”AZC²ÝÖ[ÛØ®u e‹AM”/C¢�‘†D%:_rÛƒÄ4!HBb �ñ‰ ü ²ù¿Ûmñ
|
||||||
|
õ¹>÷ÿýßÿÏsÏó/ ^�ÐÓ:ºjªüà �ôÐxiZÚí |•æ'…TŽó~l†æÍTvvâÖÍÆSd˜ôgÒ†ž¼÷Þ7ÏÂ0ñÛÓ$ð\qm#þUâ7§sÓ3ž lžx“ø²l~œbá{òßgU‘Óg
|
||||||
|
l;Þ"ý9â¥I=g,½“ü‚ø› ÷Q!_œ^~u€ûK_˜2
|
||||||
|
yµ�â»ß'¾.z<hÄft-/Ë_ l «Â]ü€Køgq§ðŽcEä0ŠÃˆa/vAÓèÁ6øʪÜE÷F׋®>W›«JüYüN¼,‹1qŸð£Pòü›¿Îoä~e·Ùev„µ2EÿŸcŸQŽw)þË8fgÃ3¶3ì¦ ‘Õ ¡ú
|
||||||
|
uëk½5ÕU•žŠòuen—(𠃲Hß",Í�ÍÉÚÄ!UYlÂÃ3&üñ†ž„Mzmb–·'L®ù‹HUüô3¹vUYy±1²¶ÀøVMŠHi=iòD%S˺”\…æ†ã> ð5'öÅ- ´J‘´én ÔMì�ûšL1üD¢IO&Äåù^S‡|‡rû ]O>ÔRlÕúîü͹ïž-òÖ6ô¨ÊW´l0ګȹA“ë™Ñ�QMU%“—µEƉáRXÏɤ0Yk4+¬µR´ÆÁŒÒ%»n³¿‰°ÍH“¹+‘0EY£LÖ2&'k¦§½QU¤Áô€ªœ‡âòE�ŠXôŠ°¨$-püœ&›L×"´a¡á¸MXùT…Tao˜Ê
|
||||||
|
BUªrIUú̺vuåÄaÍ7ÁY—K¤‡n—-¡&™£#*Èĸ9á Oªƒn·®š÷áïôy}Þ-4ø h"îÃ'\¸¯Y‡óì†x’kC%B5ÍnÀÅ„`9‘U;îë[ ÛÖ½ýq_ .—|Í[_ïÞ®Ù°¡†ÝØT]½Éšøã`®
|
||||||
|
öºU ;Íž³6å®c‚«%Z!â_FΈgïÆh�×Øy»F'îÊŒŸœ©©îý
|
||||||
|
�–Ùâ+©¾�×hÃRÄ}B¼fõpŽ'½ÅÞ¥.Kd
|
||||||
|
÷‰µˆ«ƒö�ökæm
|
||||||
|
‡¬”Èѽw¡žBxð”c[EwqÕ×r”S�aŽ—›íw0/;æ`�l>p°ˆJvÕÁ.’ÿB–LXGÜÛË uLr0‡*Öï` -,á`�l^s°ˆMÌt°‹ä·F êSÈ …4¦io·’´�h�:Fº¡Ö SOÉ KxzGI’µ;MJ¶¬Ÿ$–El-ZÑæ ¢å(Ñ;I–�»ËŽ9DùHu¨<& «*+’�q’dl�UÅ ;“A•¼Ô¡ò…Ù©L*=-m
|
||||||
|
·IÁžžnEÒôb&+
|
||||||
|
¤¨ž�-¦�R@êÏf¥˜eV”bFј*É@g°KÓ‡òQ]Šä'IÎg³Æøt&?©H{Œ’‘U5J�ÃQê¤#TGŽ*›¥º;Ixtld<Wœ
|
||||||
|
¶—Â*I§"f¤Ždu«Ë‰Ú4e/çŸÃXy£ùÉ”1µs5Ú¦ Òªƒè°÷O¥Þž§ýPí¯„` ØìR÷æKjO÷ÿeŸ"{,?oý·ýÃEk[Ä檾9ê©cÑ– �n®Ü�“çøÞã�øýÀïû£}¥
|
||||||
|
endstream
|
||||||
|
endobj
|
||||||
|
76 0 obj
|
||||||
|
1324
|
||||||
|
endobj
|
||||||
|
75 0 obj
|
||||||
|
<< /Length 77 0 R /Filter /FlateDecode >>
|
||||||
|
stream
|
||||||
|
xœk … …
|
||||||
|
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
|
||||||
|
xœ]PËjÃ0¼û+ö˜ ‚l÷ Á¡àC ÔíÈÒÊÔ’XËÿ}¥µq zÌîÌ0¬h»[çlñE^õÁX§ g¿�Bp´®¨jÐVÅ ñ&
|
||||||
|
‘Äý:Gœ:g<¼n,½„� ¾ÓgŽ´Âéªý€/ Ñäþ'i$ëF8ý¶ýÑí—þpB¡ä :ͯhßeø�‚`Ÿs§ÉÆõœäÆÏjÆÕ–Ay�s�
|
||||||
|
Iº‹K™ª�Ë[ª&»?ÍëM5u—t°Mª†Q•PYÖkwVvÉÛ8"ª…(¥ãep¬ È:<¶|È*>ÿŒÙ}
|
||||||
|
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
|
||||||
|
xœ�VkL[ç~¾sóðì€c² ç@àc°
|
||||||
|
¡$4XÖ„6!id4V8c{²1Á’EÉZEUPÕm�Ö*Š¦VʪVݦšÒ*JQ·*Z*µŠ¢þØE�¦j›ÚH™Ô Ý´^böžƒaëšvûìs¾ç}Þë÷žËwÀ TâQðˆ=t¤-Ró@õ õ{lb~NžÝú�;„ßXÝÔL:Ϲþ ¿KçNO½÷½ú,À“ž›É¤´ÉO¶Ý8Jþ?"ŸÝ"ªÞn’Lz4dòs§œ<Þ%ùC#k®0A±„Qò¿F²+¯�šÁ_ñé?#YžÖò©[ߌý°ÔSü—f
|
||||||
|
Źµs𠶆~f65s¬ Iþ€ä��`…Û¡â‘Òh©¸öE¾À/‰ âeæbÞÃM¬âEü—p?ÆÓøžÄãø.ò˜Â$F1Œñ ö#Žû±]èDl·Ú- Y|Ò;Òué’øOñŽø[ñ–xS¼"þR|L<#ž‰áÏ … Â8_â?âßçßæWù%þÛ¼�[àÚ¸&¶Êžf{Ù6ܦ*þDu¼‰ßà:Þ z®á*^ŬàøU÷.ã¹r�ÏšU>EU.áž ZÏã1œÅœÆIq¬| �På‡p�* D?Uß»Y@tǶx=5Õn—ÓQUYa·Y-’(ð Cp…®{B^<¾¨Ä§Ž©Á• qô”ŽÖdm÷ˆ9õ˜“noÑYÂõ_”l¥¿Îµ¨Áõ;®C‰/3¾1.ÊmRçi–u!¡hòär,¶x4 ±Œ@ýÈ¡¤„Fy0£[4 ‡“¿.&îñk“#G’J›ò±¿M¹M¹[¿<º6ù¥–bc¼÷öœ{?6)wum·ü-Œº`¹8 ó -;Þ? Wƒ²Î+ñƉ‰ù„’XTH¡ J¼N¯0ÖJÑê²zL“ͺõ>?aêƒ~}ÿȈ.*q
|
||||||
|
Á”xV甸^ÙR§å�L¿¼
|
||||||
|
7ĵ×ãTÄŠ[„1Ëò2Ç/Æ�iñAjXìhR���ùÔ ©î•† ©Áëj°W÷´¨X IzŸ£çc ]{šÅ‡éi¶ ´ÌÐÖ³bÎÿ-²,‰ïö¬ð A,ó-ôŠEZú¬‡ÖH8ê¸î@?'—س¥Œøð'/÷ oSFw,.qcØ‹6(G 1áë¶öÊÚJØF“ÇŽ ~è‰Þû÷ÜܵSÙØî«qVÙ[ÃŽILsñaôÒhm�îèÜÇE#]»»ØîÎŽ&2îÜÇ¢�ˆ×Á);,’…ÔÜ“g’×S»%J2S›1jüðã &UXφý•Ì'øvæ=k
|
||||||
|
×äe’Ýn�g¸ôYÁ×”óZ¬ ®iÓ¶ô¾IJ•î»¿÷
|
||||||
|
Â:yáîÅ%Ñ×D!l5~ÑçtØBBÊcÝÂ[ý[‰¨s9ì¡R
|
||||||
|
ûÇ¿)æíÓK ›—ˆ:¯ßép‘×t÷–O0(u¹�b3õ1ŒŽX0äÜ’ž¹j+9ëN&5€ NNhUä¯7Ð qLdQ§ÑÀÖ¯ê ÛìÒ^ÙÇuQGxéÞ ùV5ou=�T–>°
|
||||||
|
®>c¼ÏYí½Â»¾¸No�Ýýô’ÏãzôùŸóþ’Ë�Ï
|
||||||
|
&¼Â¾‘ÀEv(�®ÏÂ3˜âh¹\…ˆ¯1z³ž/—èý/ÑkÞgå¸ëǹ¥.Û˜³çïôþ5é7Ó½/oÌkWK° — ÑJÎlÃ�âí7(ÒŸ· ߌ¸1xîú¥pP<A×ÂÐò˜ à 탶ºFoÓõá wòFŒ1¼¾™ÇÎv—1 ;\Æ<Üìl dóÓ2QÅÞ*c‰øÉ’ 6’VM/3x˜\Æ ¬¯Œyì`#e,�ÍSe,b+ÓËX"þH €Úf‘EÌAÆ.b›i£›~ ŽC£�#‹ áÑyˆ˜ yÉ+…y“ë#Æ°ÞŒV4¥Í)Ê1OçI²l§Ø3æÊoD’i*`ºìaTeDJQŸç(–¡1ª8hfJ‘F%ïí<…™Ó³ÙtfNÞ•h–ÃÝÝ A9®³9ù@H Òr§‹™Ô|HîËåäaì(§Š©ÙùÔd¨= ‰k
|
||||||
|
Cš<X˜&E¢�Ë¥&æ²…é |05ŸÊ©qJ�ÇU7Fuäé—¥ºÛ‰\ÐÆ&òùl˜°±¸4m«92›%1•>™Ól,gÈœÓæò÷|>Œ‘w¨0�NÍîÙˆö?Ó…iÕa´™ýS黣@ýPÍ«„p(ÜŽ¨æÕîŽÿ#”y™cíŒñ}t�‡ÞÒÌ7ˆW½ÐœtT9ŸpOWd¢I!™T«:ñä«|ÏãÛZ[�4Þý?
|
||||||
|
endstream
|
||||||
|
endobj
|
||||||
|
85 0 obj
|
||||||
|
1742
|
||||||
|
endobj
|
||||||
|
84 0 obj
|
||||||
|
<< /Length 86 0 R /Filter /FlateDecode >>
|
||||||
|
stream
|
||||||
|
xœk` % ( £
|
||||||
|
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
|
||||||
|
xœ]PËjÄ0 ¼ç+tÜ=,N²=ô°JJ!‡mKÓ~€cË©¡±�â ò÷µ•�B~Œ43 m÷Ü9A¼“W=F0ÖiÂÙ/¤ +ª´UqG|«I†B$q¿Î§Î×�¥—°3ÄGúÌ‘V8=i?à4šÜ#�dݧ¯¶?ºýÂNè"”ÜC§ùí]†W9!ö¹t:‘l\/IþÇø\B͸Ú2(¯qR!I7bq+S5p{IÕd÷ó릌ú–t°Mª†Q•PY><n¨Þ�b§]“=ónŽÀj!JYy5 2dz� ²ŠÏ/àï
|
||||||
|
endstream
|
||||||
|
endobj
|
||||||
|
90 0 obj
|
||||||
|
241
|
||||||
|
endobj
|
||||||
|
91 0 obj
|
||||||
|
<<
|
||||||
|
/Type /FontDescriptor
|
||||||
|
/FontName /EAAAAC+mwa_cmr10
|
||||||
|
/FontBBox [-43 -250 1008 750]
|
||||||
|
/Flags 33
|
||||||
|
/CapHeight 683
|
||||||
|
/Ascent 750
|
||||||
|
/Descent -250
|
||||||
|
/ItalicAngle 0
|
||||||
|
/StemV 0
|
||||||
|
/MissingWidth 500
|
||||||
|
/FontFile2 92 0 R
|
||||||
|
/CIDSet 93 0 R
|
||||||
|
>>
|
||||||
|
endobj
|
||||||
|
92 0 obj
|
||||||
|
<<
|
||||||
|
/Length1 6620
|
||||||
|
/Length 94 0 R
|
||||||
|
/Filter /FlateDecode
|
||||||
|
>>
|
||||||
|
stream
|
||||||
|
xœ…Y p“ו>÷I¿$Ë–¬‡eù¡‡%ÙÖÃY¶ü�eÙ#lð ð?°±Á`[¼�Bš¤iiÚdØ”I™”RÁ¤SÚÝa“m“n¦%nšdÒ6›ÙngÙͶÛL¶“”XìùÉ„4mWã«{îã¿÷Üs¿sÎ÷Ë@ @‡�†Ðú�e•ÙÕù Ú±wÓðü S¬dýY”ÿËo¶ÏŽMSª �KXÞ›ŠoÿxþÒ úC ÕK㣃# ÿ'1 �çûDZ#cçÀvÛEãÓ{'ޖƱ=Ží©©™a\ëƒ*|þ¶g§g 7qüØ6í œ }o] Û¸>õÌìÌÜž{‡@`ø’0>»{tö›žÂ¶ o1p %äƒÂÉÞäܽ%\yŽ>Ï.°Èkð ¼o@.ÂópÎÁ“p ‚i˜„qØ
|
||||||
|
› í°¢†:B |à…20K’’ß³ï³7Ø{žÝÏÆX)ó.#¥“ô Ú@}B}Hýõ5MõRª�Ü!/“ãä!2Mlð{Üñ]ÜómÜõgp nÀËp ~ ×àE¸
|
||||||
|
—E-žƒóðex ƒÓpµ9q˜ƒÔiõt ¢ fˆ°*€�N«ÉV«²2•
|
||||||
|
¹Œ—J8–¡)®«xM¦SC§¬áí›Ü®«¥ M�‹ pvë«{ĪV¬²ÒžiÊú‹.·Ë‰ ªÔíJ}‘¡XÃWm ›ZLãƒ# Ú†µ)Á4YM#WB¡S�Ýfü\s^OG· 06SËxBbk LÀ†n³1Á6ê1Žôl춖Y?6–YïàÞο½úàÈßœÉÚÂ
|
||||||
|
w>÷pÃÇb—J¯v»^ÆcA+JžjNÐMƒ‘�°ÛÕlJÐÖðUB±MóMÖ¦SV H0ÖpNB.œWËižH„M¢Þ‰F#Ê‚˜h1&¢== ÖÆ%ˆ5<‘ ¬á„¢4Çí25�GÜ® €
|
||||||
|
Ø{/…Q‰«*„ÚdºBѧÂÖ · ÁB�݉Ð@�°ŸÛ…CMª&T·Ü.�ÛuËíjHhJÝ�útc¹†~E¬ r¿ø 3 Àš°¼À-HP– ?JObÁšG_’á£2ô1Ùû òÈpaA U®Èd ²Ðg²ðy>«z(À¹ÔÛ…^/ÏeµW%ÌôW^áØ_Õ^¥)á
|
||||||
|
-t³B÷U ·ó“Z´!Ê^•Ye3«ÌÊ”,"ç’ãlן¿a^Câèl'¦(”/# ±H8GàUB«yÒ›Á–dgÒNgƒ³Üë3ùTœÕWðVê´*�Y{�ÀäswçK2D¯7zT.¹ÛÚ,DAǽ%ú2®Ÿ &°†òr H³ Àƒ„bº`»òhà( nPÎrÖB©ªÔ~oe¥ÕPœ£Ò¯öUQv«…ÃvEÁÝï6¾|ùÄ¥^sãõïìžÍÍÅ3›+/ßá&Ï�ü«¤æ�c·ÿp4‘|íMß’¼óOféÉå§w¿úæ“ây7¡>QŸµÐjµƒ:”œ¦øÂ<Ê`s)©�©å„£u¨"ͨU”ί©æ€õ‘ö
|
||||||
|
Ãåe½ÖRL¸Öh#pL›`Ô¼´<P@¼•~_•‡rK ©ò*뉯JÐ]RO{i8«Åî«Âñ”D’ª¬NM„97ÊŸÛ²0¢ÜÜpRÞÝæêßÚ~¶”áO}½Ç šüfgý©ÙÖ¶G:zÎ xuvàåý¯9PØRñlKä§GjöŒÔf«³ø©¯µ
|
||||||
|
Ör»B!W÷ÃM¶æ©æMOLæªF¾Þ»ë{c«ŸØî�®2UFÐ.ÛçÐ.• =%9t†“WgQÚ¤‚ÐÄ© A1%„×ò¤²ÜC¯/âyÖHJ³h:ª%4GU¥m!à¤*@êi4ŽïðôTêœJ*S°WÀ ðÒÛR ííöÖïßÜs6_ªô46zÊk¦ mÛûê|û¾ëÿqëñ\Ž¡¶&Àrd }òÔc �_íßÑEÕW”56¹ÛNL¬þn¼çüÑçóÞeÙÜo$ŸÄ35 /¾Ç.€ ‚¡J#át|á,R%•‹Î’ÅQL>ð¼€s;É'“y¹ k%´Œ”£ŠEÀ{Ä£TúTUVŸ9Ûê¡ñf•D«ò
|
||||||
|
7ëUYñxN"xAÃÒ3Q T–õ~åØ?ÚJr Ÿ²ƒ¶¡‰Ç ‡^1ÚM:�<Õeáò –ÿˆÝËoã„å?
|
||||||
|
^B0_ u‡~¬à ÙåD« ŒŽL™BJ™ øÜ 5¶¡äd‹‰Õ²Ä&ê&z#âM%ØØgÓ~b«¨§Æ«õµôšë‰Í¼osÝ€Ïa÷;'\?ŽêœTY B Õý¤g&´üö€~ñê†dBЉ‚ê{wèýeôÚbð†œyT¦†·Û¥2�[2�Ó�è ¤¿À"eͤSÅ AKSn,ø± é¤[ÍZ�N0Z žI¥VIˆ9
|
||||||
|
’<x{aúÆâÖÓ½ íòw ýZÃŽõJJ.Qkó_JÆÉ’sdÞ‡ôGßÙ·øÖñ†Ùs}FÝ ®©–Ès÷<EYÛž�x;g"É9&ˆ64"—¨
|
||||||
|
¹TDC¬6 ‘%yv)Ÿ…ȉ‘bÒjè> «åºŠ30…Ñ–D´¥è‹TJq�>àU¡ÚZ‚(U Ú*…D¨í-½åOtô?»·qñv|ó—\«ý½ÍË·ú ³h�õ·ì\]T³˜œkÍ)%ûCíë þV÷¾7
|
||||||
|
Å^¼;WšÓJ.ÅûÉøÐ× øb(öò#x÷u÷î0ëÄ»¯ 3QØóyµ„0y¼N&‘å-"³O—¥EÜ÷I€á(›hâáþé»WiR¡çÁ¦–õ ‡±ÿZ¸}¿~mlçÖºFâÃfò¶ýdìÉoHnŠ XÞÑÑ㺶—L¥Z~¬ôßn8(`³}HЯ|!§œ(ˆ)
|
||||||
|
Í ^ §žHŸ:_ç ÚW„–M£ áo@T´©9Õ/ÇTG]
|
||||||
|
§%%ÿ°íìc½-IgKoMïÂW{› @j÷ìÉ/$5N]ûÈïFŽ\�'§°%èIãµÊCÅÅE%£+àuE^ ŠRÙ¦1Ñ.W©Ø®\oÞõy¨Þw À.:PÊ®�*Ĭ.Ÿ j“^Ý3÷Ó‹·ªKJª=–ÖÃC‹åÁ£Ãƒ'
|
||||||
|
YNVxMäÍxü7Ç·î(‹‹|÷Ã/m þ~|a@!—pq�˜£®Žd?“EÏ!Wõ„Š(‡Uxß%ª\™\f4h,«V¢5�tºíÙ,©Hy»Çã,ÏF j D0¦YkÑjôFâC¬Ðä$DI¥´O'’3Ž²#±–äkÙí�äò.ê·¯qF§‘t6lËM>›·v*´vW•=·DïJ&
|
||||||
|
o\¢þÇ®�F+¤ÌòÏÉGYÙ›&©þäëá™ÖáÕö*}±Ñ\…º�JÆ!¡SÈPHaê†BVRŠ‘ÐÔK²X§Sy¶Ê¬µª¼ÚHÎ�~”D¦ ûèB Rq™½ŽqÙ
|
||||||
|
5¡2KF! ßÈåÕ
|
||||||
|
©xC6ÚBm5ªxF©à È�’bMdƒBÊbàC[¤�•
|
||||||
|
Ëh$8+’F„–{P¨L�.QýKÔÈë¯/}iùÂR*G“Jò�PßýšðM9—ßHÅâj̃>¼›B1êeP=¯%{¦Œ`$ÑÛ€'ˆ¤|�6)=ÇòÕ4²#ó
|
||||||
|
”¼BÞ¤"…5Û,€F#!„¡ÏKwè9ªöÚæ]«ÍA=u™Ï`–oÜìY(uäKƒV}–”£¥ºº [I,C¦à¤Ÿ|¬+u LÆE̸“@ZîÛ�íND»³„Ã@ü€Ý…Ìä3«Ü?þ11$!&é�}$®�”(Ðî6„œf£œÎ.àõ|¦hw]–`w[Ž\Æ-ü¦l#°ÁM ÷mÞPú©Ñ+õ<ñ{?cq‚®ã B3º´DŽ¤ä:¹.5�£KKË—–’RÞ�Ë'•KÑ8ãnŒêïÆ™úó3ÿü;½ž•Ý= �b,�¡/€¾\ õЪ—žáy» ]YEŠˆƒ j8W!å·—aX§™~Òf6H\¤ÏÈÒœ’xÑÑ7ËÁÀÑ¡T › Tù‹¼):"yÀÏEOâO¥¿c¦ÈÅìŽ# ¤RV��òc/‘ŒxÕPðd«Ã[S\Rss~ãÅ –"‹¹¶|CYífÖþ¥Îö¯Œ>?
|
||||||
|
O…¶Ü8üʾ¯6tgZë‚Ž’šš’©‰ÖcÝIoqSÄqô¨gcÅ–P[Ÿgx}t(¼»á�º‡rP�ì¥ß£ÏãÛãZØ }¡˜YJmXÇkH®
|
||||||
|
ßR7 µÒRb_WœÅѤÌE †ì«0r0L)i¯Å@\˜O
|
||||||
|
WET¬YEõµ®Ó%¥G
|
||||||
|
ׂŠ£{îS3=F7!àÒ\LÿYâ`‘ˆä -àà0icä[i â¡HjE"D*SôVˆ“Õ“«Kœu[C‹
|
||||||
|
ûKçê]«ì†\»ar�Í=ðÌÆÙ-ùþ‹ûn ?·ÔÛ÷‹ss³Ï¶4?;só ËÖ—ãå5##kòÖ�_=/ÜêXÓ¿š�9[Ý
|
||||||
|
.ƒÝn°g®¶íðm>ÝVW\éßà:ï]ðùz›·¸Ý[šw4¸7�·>ÚqúôŽÈΑ `|!I¶ß9òÀÊËÍÌÆwŽ™ÒgRY<¡ûe¬†%é4u?KÙÅä)düOɸÅÞÐÒ[×K í�ÉsW¾¿ _$#bRZ_„é01wÜ‘¾+IÅ÷{w¨‹ÌûÈE!�†�L^†šX¤2ZšYd“‘DF&µjÂ*pDÌì •/ê#pÎ :0j„onZó_¢óö �
|
||||||
|
sòÝê5-õÁ²þðþÇþDù|~"y"ºü³®Ã'ÚãC;žLó"äAö ZÂu!¯ž›x9)vHeŒ4sŒÉn@ºAa:W+rˆà_
|
||||||
|
Lè=Û弿|ÅDŸI“úÏ
|
||||||
|
[*Qª<T*oêȇñŸÌÄ~pí#f¤ eø•ü©ï‰äo{›»ÎZsV÷yw<\¥Ò ýåâ�75¬"/ˆ)þ®[d Gê"¯4u µ“ÿr·Úãÿ4Û'ØußÝ’ìX
|
||||||
|
Í¡Z¿#[Í°e2>—ä‘2«I¡�FH”È,!ñ<^¶šUJø"²µVIX/Ñ‘I QsÔÚO=A%&ö45‚ƒÃ¶rF}À/Ž©4ˆtq(Ítz¶€^q‡„Ãá95æþ�_m]lªÜLþéØ7§¾7ùZ¼““*9¹*'ÛZ_4ü�–ù–én" ïmwmÿþÎÈÙÅŠM ¯Pç“ßdɸW~ ;ÞZ¿w“²Ñúö‰É L�>ÎñÒ¤M®:°»h¤u_S÷cúƬ�;ODgn-ÔWœàÄ»õ!çu#7Ë7Ô‡ª¤¤ÔŒH““R›)Z1Ûu"ñåe éRÛ�Í'„'”ŒôØ|Ž*ûôr³?u ‘¥Ý�¢;ТÛS+osøJÑÑAÊV=ÕÿátüöÁ§ÆúšÏ ^×:½ÆŒoÚxºDä¿ô;«ò¢º‡ê�¾±pà—Å“O#¦:Wo[wøL8öÏG»Vã�
|
||||||
|
¿Kž�÷@9!�)ã.• C Êt> <茗"î<AØãŽìò„ üynÑçRï+·0gdA‘ð;ƒÑÎ é"Íü:34Àvš@ÛïÿÎ ÞôÊ»(qoîv‡€_N’b~ì}_ÓÁçýlqßí}vêlúÖõ¥¡9»][ × |ýÀÖË�¶ ÷Ðþ¥ƒµ¿¿6v"`ñkü5§æ¾w >ó!ÌvòE`QxšìHö¦jæ9ØN©Q}9 ç3 b/${³>•â�éuSåätñ»Û2kÿ
|
||||||
|
¥b÷+c
|
||||||
|
/¬Ô÷žGê¥ ƒ ŒJ?‰ß¸ FKiàÞó÷ÎJÕ÷W\ù©÷! \ÅÒ�e–íl h3øè3PÍp|Ô Ûl;¨ |€;5ã(»%�ÅçbX„vmΞ›ÃâþK˜¸»€�TÔRàAu(Ý„þ´>J¸|_ÇmðÒýsð¤*-Sˆœö´LCYLË Îy&-³�An¤eûÿ g†ÇÖuñ)A& &¦´L�’Ô§eLdSZfpΣi™…\òÍ´Ìaÿmh‚˜…8ì† ƒqØ#¾E7A ÖxÚj¨Êa„9œ3…òZŒÏ&hÞ)ñåq…y±¯{„±û«Í‰Q¬Gq�yü Á™å¸v¥¸æZÜ_XÉ-(íL?!h%¬4
|
||||||
|
ÃØ3!ŽZ¬wÅ7>íhš™�ïžßc*n*1UTWW¹LáÁ¹‰)ÓZ�©mp*>7>:ï15NM™b´9Sltnt÷ü舧¼¢2<¸v¦mÐÔ2³šf¦¦F‡÷LÌìt™Ö�Î�N¹Ã¸Á4, vÛP�iÔ¿5‡é…ÁmÃÓ»+PN6{QŸA …ØèØÞ©AVÎÒ&ÖcâÙk \CزmfçØèîšôR#LµXÊD³¹¡ —�ÇZ¸ ¨ðT”UTºÛgæÝÕUÿß:"rÄϽýÂÿ:þŠH ´E^(Íd [”’IÅI»Áº†Ý²Å™áƒ3×èÚãùNçÿ’N>
|
||||||
|
endstream
|
||||||
|
endobj
|
||||||
|
94 0 obj
|
||||||
|
4925
|
||||||
|
endobj
|
||||||
|
93 0 obj
|
||||||
|
<< /Length 95 0 R /Filter /FlateDecode >>
|
||||||
|
stream
|
||||||
|
xœk`ØÃÀ°�…U<à· 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
|
||||||
|
xœ]’Ûj„0†ï÷)r¹½X4 waÊ–‚=Pۈɸj Q/|ûÆüÁB ¾Ì?N~g¢[ýTë~fÑ» eC3ëz,Mãb%±–î½>ð„©^Î�ü] "—ܬÓLC»‘¥P©Å%cч{™f»²ã£[z`ŠºmÿÍ*²½¾³ã×Ùw›Å˜ HÏ,ö{¤•F·a^Å@,òß9Õʉúy=¹ô?Åçjˆ%ž9Î GE“’¬Ðw:\c·*v}v«Ú¾þ/ž”Hk;ù-ì.ïܪ<qGqœrPŠA)(e ”{Ê
|
||||||
|
Pá© ±ÒS‘ƒÎˆ…
|
||||||
|
Pˆ ( Ô‚B= å¤<å-ˆPý ê žx ¥Á_þ
|
||||||
|
œ…Ã_Q‚à¯@=e‚¿"üå8'þðÏ8ü%pă¿Î·&ô`kÒ6lûÈÅZ×|?k¾ë[¿{MûКÑlYþúçݸ
|
||||||
|
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.
|
After 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.
+661
@@ -0,0 +1,661 @@
|
|||||||
|
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 $|w_1H_1|+|w_2H_2|$. 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
|
||||||
|
$$\label{eq:max_phase_uncertainty}
|
||||||
|
\Delta\phi(\omega) = \arcsin\left( |w_1(j\omega) H_1(j\omega)| + |w_2(j\omega) H_2(j\omega)| \right)$$
|
||||||
|
|
||||||
|
#+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]]
|
||||||
|
|
||||||
|
** 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).
|
||||||
Binary file not shown.
+560
@@ -0,0 +1,560 @@
|
|||||||
|
\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
@@ -0,0 +1,436 @@
|
|||||||
|
@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},
|
||||||
|
}
|
||||||
Reference in New Issue
Block a user