Iterative Feedback Tuning with Application to Robotics

Size: px
Start display at page:

Download "Iterative Feedback Tuning with Application to Robotics"

Transcription

1 ISSN ISRN LUTFD/TFRT SE Ieave Feeback Tnng wh Applcaon o Robocs Alessano Bn Depamen of Aomac onol Ln Inse of Technolog Decembe 3

2

3 Depamen of Aomac onol Ln Inse of Technolog Box 8 SE- Ln Sween Ahos Alessano Bn Docmen name MASTER THESIS Dae of sse Decembe 3 Docmen Nmbe ISRN LUTFD/TFRT SE Spevso Rolf Johansson an Anes Robesson a Depamen of Aomac onol n Ln. Eoao Mosca a Unvesá Fenze. Sponsong oganzaon Tle an sble Ieave Feeback Tnng wh Applcaon o Robocs. Ieav mnng av nfomaonsåeföng me obokllämpnng. Absac Man conol objecves can be expesse n ems of a ceon fncon. Geneall, explc solons o sch opmzaon poblem eqe fll knowlege of he plan an sbances an complee feeom n he complex of he conolle. In pacce, he plan an he sbances ae selom known, an s ofen esable o acheve he bes possble pefomance wh a conolle of pescbe complex sch as fo example a PID conolle. The opmzaon of sch conol pefomance ceon pcall eqes eave gaen-base mnmzaon pocees. The majo smblng block fo he solon of hs opmal conol poblem s he compaon of he gaen of he ceon fncon wh espec o he conolle paamees: s a fal complcae fncon of he plan an sbance namcs. When hese ae nknown, s no clea how hs gaen can be compe. Ieave Feeback Tnng IFT s a npop aa-base esgn meho fo he nng of esce complex conolles. I oes no epen on he plan moel, lzes I/O aa onl. Theefoe IFT s obs agans he plan moel ncean. A each eaon, an pae fo he paamees of he conolle s esmae fom aa obane pal fom he nomal opeaon of he close loop ssem an pal fom a specal expemen. No enfcaon pocee s nvolve. In hs hess nng of obo jon conolles sng IFT s consee. The ffeen IFT-schemes have been vefe n smlaon an n eal expemens on an nsal obo manplao ABB Ib-. Kewos lassfcaon ssem an/o nex ems f an Spplemena bblogaphcal nfomaon ISSN an ke le Langage Englsh Sec classfcaon Nmbe of pages Recpen s noes ISBN The epo ma be oee fom he Depamen of Aomac onol o boowe hogh: Ln Unves Lba, Box 3, SE- Ln, Sween. Phone 46 46, Fax

4

5 3 To m faml

6 4

7 onens Acknowlegmens.. 7. Inocon.. 9. IFT Descpon 3. The Nonlnea ase Mofcaons an Impovemens o IFT Applcaons on an Insal Robo 7 6. onclsons Bblogaph

8 6

9 Acknowlegmens I wol lke o hank Pofesso Rolf Johansson fo he welcome, he avalabl an he me ha he gave me ng hs wok. Hs hman an pofessonal wll be wh me foeve. I wol lke o hank D. Anes Robesson fo he sppo an connos help ha he gave me fo he heoecal pa an moeove fo he as an nghs n whch we woke ogehe n he obo laboao whee we also fomlae a heo smmaze as he bes esls come beween mngh an fve n he monng. He s a peson ha caes abo eveone an I wll neve foge hm. I wol also lke o hank m Ialan spevso Pofesso Eoao Mosca fo hs avce o se o on hs expeence n Ln an also fo hs kn avalabl ng hese monhs. I am also gaefl o all he people of he epamen of Aomac onol a Ln Inse of Technolog fo he help an sppo. 7

10 8

11 . Inocon Man conol objecves can be expesse n ems of a ceon fncon. Geneall, explc solons o sch opmzaon poblem eqe fll knowlege of he plan an sbances an complee feeom n he complex of he conolle. In pacce, he plan an he sbances ae selom known, an s ofen esable o acheve he bes possble pefomance wh a conolle of pescbe complex. Fo example, one ma wan o ne he paamees of a PID conolle n oe o exac he bes possble pefomance fom sch smple conolle. The opmzaon of sch conol pefomance ceon pcall eqes eave gaen-base mnmzaon pocees. The majo smblng block fo he solon of hs opmal conol poblem s he compaon of he gaen of he ceon fncon wh espec o he conolle paamees: s a fal complcae fncon of he plan an sbance namcs. When hese ae nknown, s no clea how hs gaen can be compe. The conbon of [Hjalmasson, Gnnasson, Geves, 994] was o show ha an nbase esmae of he gaen can be compe fom sgnals obane fom close loop expemens wh he pesen conolle opeang on he acal ssem. Fo a conolle of gven pcall low-oe sce, he mnmzaon of he ceon s hen pefome eavel b a Gass-Newon base scheme. Fo a wo-egee-of-feeom conolle, hee bach expemens ae o be pefome a each sep of he eave esgn. The fs an h smpl conss of collecng aa ne nomal opeang conons; he onl eal expemen s he secon bach whch eqes feeng back, a he efeence np, he op mease ng nomal opeaon. Hence he aconm Ieave Feeback Tnng IFT gven o hs scheme. Fo a one-egee-of-feeom conolle, onl he fs an secon expemens ae eqe. No enfcaon pocee s nvolve. As n an nmecal opmzaon one, a vaable sep sze can be se. Ths allows one o conol he ae of change beween he new conolle an he pevos one an s an mpoan aspec fom an engneeng pespecve. Fhemoe a vaable sep sze s he ke o esablshng convegence of he algohm ne nos conons. Wh a sep sze enng o zeo appopael fas, eas fom sochasc aveagng can be se o show ha, ne he conon ha he sgnal eman bone, he acheve pefomance wll convege o a local mnmm of he ceon fncon as he nmbe of aa ens owa nfn. The opmal IFT scheme of [Hjalmasson e al, 994] was nall eve n 994 an pesene a he IEEE D 994. The IFT meho s appealng o pocess conol engnees becase, ne hs scheme, he conolle paamees can be sccessvel mpove who eve openng he loop. In aon, he ea of mpovng he pefomance of an alea opeang conolle, on he bass of close-loop aa coespons o a naal wa of hnkng. 9

12

13 . IFT Descpon. Two-egee-of-feeom conolles an IFT The heo n hs secon s base on [Hjalmasson, Geves, Gnnasson, Leqn, 998]. We conse an nknow e ssem escbe b he scee moel G v whee G s he lnea me-nvaan opeao, s he mease op, s he conol np, {v } s an nmeasable pocess sbance an epesens he scee me nsans emak: o eas he noaon he me nex s somemes lef o. We conse ha hs ssem s o be conolle b a wo-egee-of-feeom DOF conolle: whee an ae lnea me-nvaan ansfe fncons paameze b n some veco R, an { } s an exenal eemnsc efeence sgnal, nepenen of {v }. Noce ha s possble fo an o have common paamees. v G - Fg.. A wo-egee-of-feeom conolle sce We wll se he noaon an fo he op an he conol np of he ssem n feeback wh he conolle, n oe o make he epenence of hese sgnals on he conolle paamee veco explc.

14 Le be a ese op esponse o a efeence sgnal fo he close loop ssem. Ths esponse ma be efne as he op of a efeence moel T sch ha T 3 b fo he IFT meho knowlege of he sgnal s sffcen. The eo beween he acheve an he ese esponse s v G G G ˆ ~ So, sng 3: v G T G G ˆ ~ Ths eo consss of a conbon e o ncoec ackng of he efeence sgnal an an eo e o he sbance v. Fo a conolle of some fxe sce paameze b, s naal o fomlae he conol esgn objecve as a mnmzaon of some nom of ~ ove he conolle paamee veco. We wll conse he followng qaac conol pefomance ceon: 4 whee N s he nmbe of samples consee an E[ ] enoes expecaon wh espec o he weakl saona sbance v. The fle L s a feqenc weghng of he eo beween he ese esponse an he acheve esponse. The fle L weghs he penal of he conol effo. The can of cose be se o, b he gve ae flexbl o he esgn. The opmal conolle paamee * s efne b: mn ag * J 5 The objecve of he ceon 4 s o ne he pocess esponse o a ese eemnsc esponse of fne lengh N n a mean sqae sense. As fomlae, hs s a moel efeence poblem wh an aonal penal on he conol effo. N N L L E N J ~ λ

15 3 Le T ans S enoe he acheve close loop esponse an sensv fncon wh he conolle {, }: G G T G S. Gven he nepenence of an v, J can be wen as: The fs em s he ackng eo, he secon em s he sbance conbon, an he las em s he penal on he conol effo. RITERION MINIMIZATION We examne now he mnmzaon of he ceon 4 wh espec o he conol paamees veco fo a conolle of specfe sce. To oban he mnmm a necessa conon s ha he fs evave of J w... he conolle paamee s zeo: 6 whee fo smplc we assme L L. If he gaen J col be compe, hen he solon of he pevos eqaon wol be obane b he followng eave algohm: J R γ 7 Hee γ s a posve eal scala ha eemnes he sep sze an R s some appopae posve efne max noce also ha, as we wll show lae, γ { } { } [ ] N N L E N v S L E N T L N J λ N N N E N J ~ ~ λ

16 ms obe some consans fo he algohm o convege o a local mnmm of he cos fncon J. R eemnes he pae econ an s heefoe ccal fo he pefomance of he algohm. Tpcall we choose a Gass-Newon appoxmaon of he Hessan of J, so: R N N T λ T 8 We wll see ha all he sgnals esmaes neee n hs expesson of R wll be avalable fom he IFT algohm. As sae he poblem s nacable snce nvolves expecaons ha ae nknown. J Howeve, sch a poblem can be solve b eplacng he gaen b an nbase esmae. In oe o solve hs poblem, one nees o geneae he followng qanes: he sgnals ~ an ; ~ he gaens an ; 3 nbase esmaes of he pocs ~ ~ an. These qanes can be obane b pefomng expemens on he close loop ssem fome b he acal ssem n feeback wh he conolle,. { } We wll see ha 8 can be expesse s follow: R N es es N T λ es es T. hee an n he seqel, es an an es enoe he esmaes of 4

17 5 OMPUTATION OF THE GRADIENT Nong ha ~, we can we: [ ]. ~ v S T T T v G G G G G G Fom he expesson above we see ha he gaen epens also on he no compable qanes T an S snce he epen on he nknown ssem. Theefoe, nless an accae moel of he ssem s assme o be avalable, he sgnal can onl be obane b nnng expemens on he acal close loop ssem. Noce ha he las wo ems n he expesson above nvolve a oble fleng of he sgnal an v hogh he close loop ssem: [ ]. T v S T T So we can we: T T T T The em T - can be obane b sbsacng he op sgnal fom one expemen of he close loop ssem fom he efeence an b sng hs eo sgnal as efeence sgnal n a new expemen. In each eaon of he conolle nng algohm, we wll se hee expemens, each of aon N, wh he fxe conolle { }, ˆ opeang on he acal plan. We wll see ha onl he secon expemen s a specal expemen, he fs an he h js conss of collecng aa ne nomal opeang conons. We enoe he N-lengh efeence sgnals b j, j,,3, an he coesponng op sgnals b j, j,,3.

18 6 We have: 3 So he expessons fo he op sgnal ae: 3 3 v S T v S T v S T Hee j v enoes he sbance acng on he ssem ng expemen j a eaon hese sbance can be assme o be mall nepenen snce he come fom ffeen expemens. Wh hese expemens ~ s a pefec ealzaon of ~ whee he sbscp enoes he eaon nmbe, whle ˆ 3 es s a pebe veson b he sbance v an 3 v of. So now we have an esmae of he gaen ha can be se n 6 an so n 7 fo he pang conol paamee law. In an analogos wa, we can oban an esmae of he gaen. Fom [ ] v S v G G o we can see ha he hee expemens escbe above geneae he coesponng conol sgnals: [ ] [ ] [ ] 3 3 v S v S v S

19 7 These sgnals can smlal be se o geneae he esmaes of he np elae sgnals eqe fo he esmaon of he gaen 6. Inee s a pefec ealzaon of, whle ˆ 3 es Resmng, wh hese wo gaen esmaes an expemenall base esmae of he gaen of J can be fome b akng: N N es N es N J es ~ λ

20 8 ONE-DEGREE-OF-FREEDOM ONTROLLERS In he case whee he smplfe conolle sce ˆ s se,.e.,, he algohm smplfes becase he h expemen becomes nnecessa. Theefoe, n he case of a one-egee-of-feeom conolle, he fs wo expemens ae n wh he same efeence sgnals as he case of a wo-egeeof-feeom conolle,.e., an he gaen esmaes ae obane b: es es G - v Fg.. A one-egee-of-feeom conolle sce

21 .. Two-egee-of-feeom conolles EXAMPLE onse he plan an he conolles G s s s.6535 q q.8469 q.6854q.39 q.8469 whee he opeao q s he fowa shf opeao, ha s qf f. We a a sbance v as n gven b a nose n whch s he op seqence of whe Gassan nose wh zeo mean an vaance. flee b he followng feqenc weghng fncon W s s s The ppose s o mpove he sep esponse ackng poblem. We ake he followng paamezaon of he wo conolles: q.8469 q 3 q.8469 so ha we have hee paamees,,, 3 o ne, sang fom: We se he followng qaac ceon:.39 3 J N 9

22 an he same fxe sep sze fo all he paamees, γ., an a Gass-Newon appoxmaon of he Hessan of J fo he max R. Pefomng he IFT algohm escbe above we ge, afe 5 eaons, he fnal paamees an he followng esls: Nmbe of Vale of he fnal eaons ceon J nal vale of he ceon.8443 Fg..3 Fg..4 Sep Response sep a, ample Doe lne: nal - Sol lne: fnal

23 .. Two-egee-of-feeom conolles EXAMPLE We now appl he IFT scheme o he nng of he conolle fo a flexble sevo, a wo-mass-ssem wo masses connece wh a spng. The followng fge show he skech of he pocess. Fg..5 We have he wo masses m an m. The spng beween he masses has he spng consan k. The vscos ampng coeffcens ae an, especvel. One of he masses s ven b a D-moo. Hee we neglec he nenal namcs of he moo. The foce fom he moo s popoonal o he volage, ha s: k F m Foce balance eqaons gve he followng namcal moel: p p k p p m F p p k p p m Inocng he sae veco T p p p p x ] [ an p, he ssem can be wen on sae-space fom, x B Ax x whee m m k m k m k m m k A, m k B m, k k

24 We se he followng consans an coeffcens: m.9 kg m.44 kg 3. N/m/s 3.73 N/m/s k 4 N/m k m.96 N/V k 8 V/m These vales coespon o he eal pocess whch wll be consee n nex secon. The ppose s o conol he poson p. Assmng ha we col mease all he saes, sng a sae feeback conol saeg, we col place he poles fo he close loop ssem feel conollabl max has fll ank b sng he conol law: Lx l b popel choosng he paamees L [l l l 3 l 4 ]. Hee s he efeence vale an l s a scala gan affecng he oveall gan an s chosen o ge he coec saona gan. We choose he followng ese poles fo he close loop ssem c.l.s.: b b, 3,4.4 ± 7.6 ± j5.7 j4.9 Fg..6 Noe he pool ampe fs pole-pa, whch we have chosen on ppose.

25 The gve an oscllao esponse fo he poson of he secon mass an o am wll be o mpove he coesponng conolle o ge a bee behavo. Assme now ha onl he poson p s measable. Ths we ae face wh a conol poblem whee we wan o feeback sgnals whch we can no mease. We nee o esmae he pocess sae veco sng an obseve Kalman fle escbe b: xˆ Axˆ B K xˆ whee xˆ enoes he saes of he obseve. K can be chosen so ha he obseve saes ae appoachng he eal saes wh an abal fas ae of convegence sall sch ha he obseve namcs wll be.5 o mes fase han he close loop ssem. We ge: L [ ] K We wll se he obseve saes n he feeback law nsea of he eal saes whch we can no mease. The Smlnk scheme of he conolle ssem s shown below. Fg..7 B ecomposng he sae-space ealzaon we can easl fn he ansfe fncons fom o an fom o, so ha we col have a wo-egee-offeeom conolle sce as below. Fg..8 3

26 The nal feefowa conolle s: s.36 s 47.7 s 99s s 5.8 s 977 s 548 s 6944 The nal feeback conolle s: 3 33 s. s 835 s s 5.8 s 977 s 548 s 6944 These nal conolles gve he followng esponse poson secon mass: Fg..9 Inal sep esponse The paamezaon chosen o be se fo he IFT algohm s: 4 s s s 4 s 5 4 q s 3 3 s 3 q 3 q 4 s 6 3 s s 3 7 q s 8 9 s 3 Noce he chose o keep he same enomnao n he wo conolles an. Ths o keep he sce e b he obseve an he sae feeback conol esgn. 4

27 We can now appl he IFT algohm o pae he conol paamees of he feeback an he feefowa conolle. We wll se he followng cos fncon: J N N As wen above we nee hee expemens an we wll mplemen hem sng Smlnk. Acall n hs example we o no se he sbance v so we o no nee he h expemen. The fs an he secon expemens have he followng Smlnk schemes: Fg.. Fs IFT expemen Fg.. Secon IFT expemen: - The ese op s chosen as: a.5 T s a.5 whee a3.5 an as a n sep. 4 5

28 Fg.. Dese op Sang fom an nal ceon J.497 an sng a small sep sze of γ.5,, so ha we can speak n ems of fne nng, an a Gass-Newon appoxmaon of he Hessan of J fo he max R, afe n 6 eaons, we ge he followng fnal paamees: an he fnal ceon vale J 5 n

29 Fg. -3 Sep Response sep a, ample Sol lne: fnal afe 6 eaons Dashe lne: nal Doe: ese Ieaon nmbe Fnal os Fg..4 Fnal vale of he cos fncon J vess he IFT eaon nmbe Inal cos:

30 Fg..5 Lef: afe eaons Rgh: afe eaons Fg..6 Lef: afe 3 eaons Rgh: afe 4 eaons Fg..7 Lef: afe 5 eaons Rgh: afe 6 eaons 8

31 The paamees pang s shown n he followng fges. Fg Ieaon nmbe Fg..9 Fg.. 9

32 Fg.. Fg.. Fg..3 3

33 Fg..4 3

34 THE REAL PROESS AND THE PROBLEM OF FRITION An mplemenaon on he eal pocess has been e b a elevan poblem has been enconee: fcon. In hs case, a nonlnea conbon fcon oes no allow he IFT meho o wok popel an n o specfc case, he fcon was oo mpoan an he calclaons of - ng he eaons of he algohm wee soe. Fg..5 In he followng pce, we show he behavo of he ssem an we can nesan how n hs case he mplemenaon of IFT was almos mpossble becase we col nehe check he esl of he fnal conolle no keep sabl popees. Fg..6 Poson esponse o a sqae wave efeence sgnal A poblem wh fcon an a pool ne nal conolle s ha he pocess wll sck a posons whch ae fa fom he efeence vale an whch ma change a lo fom one eaon o anohe. The expemens on he eal pocess, even we ae n pesence of a lm case of hge fcon, showe he lm of IFT ha sall s base on he measemens of he eo beween he op of he ssem an he ese esponse. Howeve we wll show n he nex chape how s sll possble o appl he IFT meho o he case of eemnsc nonlnea ssems. 3

35 . onvegence of IFT onse a scee me lnea me-nvaan LTI moel v sbance G v an le be he ssem be conolle b he conolle wh R. n In hs secon we sae exac conons fo whch he conolle paamees pae wh he IFT algohm convege o he se of saona pons of he ceon Le D be a convex compac sbse of N J E N. n R. We noce he followng conons on he nose, he conolle, he close loop ssem an he sep szes of he algohm, especvel. V In an expemen, he sgnal seqence v,,, N consss of zeo mean anom vaables whch ae bone: v fo all. The consan an he secon oe sascs of v ae he same fo all expemens, whle seqences fom ffeen expemens ae mall nepenen. Thee exss a neghbohoo Θ o D sch ha s wo mes connosl ffeenable w... n Θ. All elemens of he ansfe fncons have he poles an zeos nfoml bone awa fom he n ccle on D. 33

36 S The lnea me-nvaan close loop ssem s sable an has all s poles nfoml bone awa fom he n ccle on D. A The elemens of he seqence { } A The elemens of he seqence { } γ sasfes γ an γ. γ sasfes γ <. Theoem Hjalmasson, 998 J onse he IFT algohm gven b γ R. Assme ha V,,, S, A an A hol. Assme ha R s a smmec max whch s geneae b he expemens a eaon an sasfes I R δi δ fo some δ >. Then lm ˆ : J ' D c { } on a se A { D }. The basc eqemen fo convegence s ha he sgnals eman bone hogho he eaons snce he esl onl apples o he se A noce n he heoem. The powe of he heoem s ha apa fom he assmpon of lnea an menvaance hee ae no ohe assmpons on he popees of he ssem. The same hols fo he conolle: he complex of he conolle s aba an he esl hs apples o smple PID conolles as well as o moe complex ones. I s also mpoan o noce ha even hogh he sbances have o have he same secon oe sascs fom expemen o expemen, s no necessa ha he sbances ae saona ng one expemen. 34

37 .3 The sep sze: a ccal choce In hs secon we scss an show wh some expamples, how ccal s he choce of he sep sze γ n he paamee pae law Le s ake he followng obo moel: J γ R. G s s s s 5s conolle b a PD conolle: wh N. G s k p st T s N Fg..7 To mplemen IFT on hs ssem we se he ceon J N N an a Gass-Newon appoxmaon of he Hessan fo he max R. We se as ese op wh as he n sep an a4. T a s a Sang wh he nal paamees k p an T we ge he an nal cos J.77 an he followng esponse: 35

38 Fg..8 Response wh he nal PD paameesfll an he ese op oe Pefomng 5 eaons of IFT sng a sep sze γ. we ge a fnal cos J 5.43 an he followng esponse: Fg..9 Fnal esponse afe 5 eaons of IFT sng a sep sze of.fll, nal esponseashe an ese op oe 36

39 We can look a he cos fncon vales as fncon of he eaons nmbe: Fg..3 sep sze γ., fnal cos J 5.43 We pefom now he same nmbe of eaons, sang fom he same nal paamees b sng a bgge sep sze, γ.5. The esponse s: Fg..3 37

40 We ge he same fnal cos vale J 5.43 b f we look a he gaph of J as fncon of he eaon nmbes we can obseve an neesng behavo: Fg..3 sep sze γ.5, fnal cos J 5.43 We can see ha afe 3 eaon we ge a lowe, even f no conseabl, cos fncon vale J 5.4. So s no sefl o conne n he eaons an s enogh o sop he IFT algohm js afe 3 eaons. Ths s an mpoan aspec of he eave nng mehos snce can happen ha we fn a local mnmm js afe few eaon an f we on look a he cos vale ng he eaons can happen ha we can ge lowe pefomance fom he paamees pang. Ths s e n man pa o he egla sface of he cos fncon J an so s ccal o choose a gh sep sze o a leas a goo sop le fo he algohm. Anohe mpoan sse s he appoxmaon of he gaens an he fom of he max R. In fac, n case of ve egla sface of J, we col nee o se a ve small sep sze f he appoxmaon s no accae. Ths hols o pefom a hgh nmbe o eaons fo he IFT n oe o ge mpovemens of he conolle. 38

41 Ineesng s also he behavo of he IFT mplemenaon afe 5 eaon an wh a sep sze eqal o.9: Fg..33 sep sze γ.9, fnal cos J 5.43 In hs case we can see ha we fn moe han one local mnmm ng he IFT mplemenaon o a leas fo ffeen eaon nmbes n whch we col sop nsea of avng a 5 eaons. 39

42 4

43 3. The Nonlnea ase In hs chape we scss an analss of he popees of IFT when apple o nonlnea ssems conolle b a ne a lnea conolle. We wll assme ha he ssem o be conolle s gven b he followng nonlnea sae-space moel: x f x,, w h x, v * whee f f x,, w an h h x, v ae smooh fncons, whee x epesen he sae veco a me, whee an ae he scala nps an ops an whee w an v ae exenal sbances. We wll also assme ha he ssem s conolle b he followng lnea menvaan conolle q,: q, whee s he exenal efeence sgnal an s he paamees veco. Poceng n an analogos wa of he pevos secons, we ge, ffeenang he ssem eqaons w : whee x' f ' h x x' f x x' ' ' ' ' ' ' x' ' x ' f f x,, w f x f x,, w x f f x,, w 4

44 4 As he lnea case, f follows ha he gaens an can be obane b fs pefomng a smlaon sng he ssem * wh as efeence sgnal an collecng he sgnals,, x an w,,,n whee he sbscp enoe ha he sgnals sem fom he fs smlaon sng ssem *. Wh hese sgnals a han, n a secon smlaon se he lnea me-vang feeback ssem s x H B x A x wh efeence sgnal,, q q s an wh he specal choce of me-vang maces,,,,, v x h x H w x f B w x f x A whch ae fncon of he sgnals x,, w, v n he fs expemen. I hen hols ha an x x.

45 43 B epeang he secon smlaon wh a new efeence sgnal,, q q s j he e gaens wh espec o j can be obane. To oban he gaen wh espec o he complee paamee veco n R, he secon smlaon has o be pefome n mes. A awback wh hs meho s ha he nmbe of expemens s popoonal o he nmbe of paamees ha ae o be ne. An alenave appoach o avo hs s sggese b Sjöbeg an Agaval n [Sjöbeg e al, 997] an b De Bne, Aneson an Geves n [De Bne e al, 996] whee a lnea-vang moel s enfe. Anohe appoach fo he conol of non lnea ssems s he evelopmen of IFT e o Hjalmasson e al 998, ha we now escbe. Fo each eaon n J R γ he IFT meho ses wo expemens, each of aon N sa, wh he fxe conolle opeang on he acal plan. Noce ha cona o he meho olne above he nmbe of expemens s fxe o wo egaless of he menson of he paamee veco. When - s se as efeence n he secon expemen n IFT, he ssem eqaons n he secon expemen can be wen:,,, v x h w x f x Appoxmang he fs wo eqaons b a fs oe Talo expanson aon x,, w,, v, he close loop ajecoes n he fs expemen wh as efeence sgnal, gves x h f x f x x x δ ** whee w f f x f f w f w x w v h x h h v h v x v δ

46 Above he sppesse agmens ae sgnals fom he fs expemen: f f x,, w. Noce ha an δ can be egae as exenal sgnals snce he ae fncons of vaables n he fs expemen an w an v onl. Dsegang hese sgnals we see ha he fs oe appoxmaon s encal o he lnea me-vang eqaons seen pevosl ha geneae he e gaen excep fo he fac ha he efeence sgnal - n he expesson of s flee b ' wheeas n ** s no. The eason of hs las ffeence s ha n he IFT algohm hs fleng s one afe an no befoe he secon expemen. These smlaes sgges ha shol be possble o mpove he pefomance fo nonlnea ssems sng he sana IFT pocee ne he followng conons: - he fs oe Talo appoxmaon s easonabl accae; - he sgnals an δ ae small compae o f x x f an x, especvel; h x - he eo e o commng he' opeao an he close loop ssem s small. I ma seem as f hese sae conons ae qe escve. Howeve, pacce, as wll be evence below, has shown ha hs oes no seem o be he case fo man ssems. One eason fo hs s ha sffces o be able o compe a escen econ, he exac gaen s no necessa. Howeve, mgh be necessa o ece he sep-sze γ when a pebe gaen esmae s se. Fhemoe, cae has o be execse f e.g. a Gass- Newon pae s se snce he jon effec of he gaen pebaon an he mofcaon of he seach econ case b R s ave an one ma en p n an ascen econ. We concle ha fo nonlnea ssems, mgh be wse o se a small sep-sze. A SIMULATION EXAMPLE We wll conse now he followng nose-fee nonlnea ssem whch has x an z as saes: x x.x z z.x z 3 3..z 3 44

47 45 Remembeng ha he sbscps enoe expemen nmbe n he IFT pocee, fo hs ssem he gh-han se of he fs eqaon of ** becomes: ˆ z z x x x x f x f x γ ξ an.4.4. ˆ z x x ψ λ We ll conse he ssem conolle wh he PI conolle. q. We can pove ha ξ omnaes λ an also γ ψ <<. Ths ncaes ha he pebaon ψ λ s no ve sgnfcan. So we can sa ha f x f x x s a easonable appoxmaon of he secon expemen. Ths ncaes also ha, pove he commaon ' an he close loop ssem oes no nflence he sgnals oo mch, shol be possble o se he IFT on hs ssem. We choose a efeence sgnal as a peo of a sqae-wave wh peo me 5. Fg. 3. Refeence sgnal fo he fs IFT expemen The ese op s aken o be he efeence sgnal of he fs IFT expemen an flee hogh he followng low-pass fle.95.5 q q T

48 46 Fg. 3. Dese esponse The paamees,, 3, 4 ae ajse n he followng scee me genealzaon of a PID conolle sce: q q q q We se fo he IFT algohm: - a cos fncon N J - a Gass-Newon appoxmaon of he Hessan fo he max R - a sep sze γ.5. The nal paamees ae:. 4 3 an he coesponng esponse can be seen n he followng fge.

49 Fg. 3.3 lose loop esponse wh he nal PID paamees fll an ese esponse oe As can be seen he ssem exhbs some qe nonlnea behavos. Applng he IFT algohm, we oban he followng esponse afe 5 eaons: Fg. 3.4 Responsefll afe 5 eaons, ese esponse oe an esponse wh he nal PID paamees ashe 47

50 The ceon eceases monooncall ng he eaons as shown below n he plo an n he able he sbscp enoes afe how man eaon he vales sem fom. Fg. 3.5 os J as fncon of he nmbe of IFT eaons eon vales J nal 56.3 J J 6.3 J J 4.7 J 5fnal.496 A compason wh he nal esponse shows ha he IFT has manage o mpove he pefomance conseabl. I shol also be noe ha he smple lnea conolle makes a spsngl goo job on hs nonlnea ssem. The coesponng fnal conolle s gven b: q 8.999q -.447q 5. q 48

51 4. Mofcaons an Impovemens o IFT 4. Mofe ceon One of he feqen paccal se of conolle esgn s o ne a conolle of fxe sce fo example a PID conolle n sch a wa ha he sep esponse of he close-loop ssem has a mnmal selng me wh a small oveshoo. The objecve n sch applcaons s o move he op of he close-loop ssem qckl fom one efeence vale o anohe; howeve, he pacla shape of he ansen esponse fom he nal efeence vale o he fnal vale s of no mpoance, pove ha oes no have lage oveshoo. In aon, who knowlege of he acal ssem whch s a majo eason fo sng IFT s no known n avance how fas a selng me can be acheve fo hs pacla ssem wh hs pacla conolle sce. B mposng he ene esponse of he close-loop ssem hogh a specfc choce of a ese esponse, ahe han js he enpon of hs ansen esponse, he classcal IFT ceon leas o conolle paamees ha ealze a compomse beween fng he ansen esponse an fng he new efeence vale, even hogh he se oes no cae abo he exac shape of he ansen esponse. Insea, b mposng a mask on he ansen esponse, he opmzaon wll ne he conolle paamees n sch a wa as o acheve he new ese efeence vale who focsng on a pacla pe-mpose ansen esponse ha s pehaps no naall acheve b he close-loop ssem. We can noce a vaan of he conol pefomance ceon 4 n whch he sgnals ~ an ae me weghe b weghngs w an w, especl. Ths he ceon N N J E L ~ λ L N s eplace b N N J E w L ~ λ w L N whee w an w ae an nonnegave nmbes. The flexbl offee b he me weghngs w an w s ha he allow one o p ffeen weghngs on ffeen pas of he me esponses. A paclal neesng applcaon s when zeo weghngs ae p on he ansen esponse of he op esponse o a sep change n he efeence sgnal. 49

52 5 In hs case he ceon becomes: N N m L L E N J ~ λ an we sa ha a mask of lengh s p on he ansen esponse of he ackng eo. B mposng a mask on he ansen esponse one oes no wase he avalable egees of feeom n he conolle paamees on he machng of a specfc an enel aba ansen esponse. Insea one can focs hese paamees enel on achevng a fas selng me. The cos acheve afe he maske neval s alwas smalle han when no mask s se. 4.. Smlaons sng weghe IFT algohm Impovng he selng me onse he plan. s s s P One wsh o ne a PID conolle n oe o acheve a bee selng me fo he close loop ssem. onse he sana fom of he PID conolle: s T T s k s G p PID ha fo he phscal ealzaon we change n: T s s s T k s G p τ wh τ an he followng nal PID paamees:.5 p T T k so ha he nal conolle s: s s s s s G 3 τ

53 Ths els he ve slggsh esponse shown n he fge below. Fg. 4. lose loop sepample esponse wh nal PID paamees We se n he IFT algohm a Gass-Newon appoxmaon of he Hessan fo he max R, a sep sze of. an he followng ese esponse : T s.5.5 Fg. 4. Dese esponse o a sep of ample The applcaon of he classcal IFT ceon sng he cos fncon J N N 5

54 els, afe 5 eaons, he esponse: Fg. 4.3 lose loop sep esponsefll obane wh he classcal IFT ceon an sng he ese esponseoe Dashe cve: esponse wh he nal PID paamees. Ths esponse s ve nsasfaco; hs s n lage pa e o an nfonae choce of nal paamees. Wh he se of a fxe mask of lengh secons, he mnmzaon of he mofe IFT ceon wh he same nal paamees leas o he followng close-loop esponses, obane afe 5 an 3 eaons especvel. Fg. 4.4 Sep esponsefll afe 5 eaonslef an afe 3 eaonsgh sng a mask of lengh, he ese esponseoe an he sep esponse wh he nal PID paameesashe Ths esponse s bee han ha obane wh a efeence ajeco. 5

55 Fnall a mask of eceasng lengh s se, wh an nal lengh of secons, an wh he same nal paamees agan. A eve eaon of he IFT scheme, he lengh of he mask s ecease b secons. Fg. 4.5 Sep esponsefll afe 5 eaonslef an afe 3 eaonsgh sng a mask of eceasng lengh, he ese esponseoe an he sep esponse wh he nal PID paameesashe Obseve he amac mpovemen of he esponse e o he se of a mask of eceasng lengh, leang o a seqence of cos cea ahe han a one-sho ceon, an o a ffeen seqence of paamee vecos han esle wh he ec se of a mask of lengh secons. Resmng, we can se he ceon J N N o compae he esls. The nal cos s J.94 an wh he mplemenaon of he classcal IFT ceon we ge J 5.. Usng IFT wh fxe mask Usng IFT wh mask of eceasng lengh J 5.69 J 5.34 J 3.7 J Fg. 4.6 The sbscp enoes afe how man eaons he vales sem fom. The nal paamees ae:

56 Afe 5 eaon of he classcal IFT ceon: Usng he mofe ceons: IFT wh fxe mask IFT wh mask of eceasng lengh.538 ; ; ; ; ; ; Fg

57 4.. Smlaons sng weghe IFT algohm Impovng he oveshoo In hs secon we wll pesen wh an example how he mofe ceon sng me weghngs can mpove he esponse pefomance n ems of maxmm oveshoo. onse he plan scee me:.493z.95 P z 3 z.743z an he scee conolle: G.6z.5z.3 z z.68z.39 Ths els a sep esponse wh a conseable oveshoo shown below: Fg. 4.8 Response o a sep of ample a wh he nal conolle paamees Usng n he classcal IFT algohm: - a paal paamezaon of he conolle nmeao - a cos fncon J N - a Gass-Newon appoxmaon of he Hessan fo he max R - a sep sze γ. 55

58 we ge, especvel afe 4 an 9 eaons, he followng sep esponses. Fg. 4.9 Sep esponsefll afe 4 eaons of he classcal IFT algohm an sep esponse wh he nal paameesoe Fg. 4. Sep esponsefll afe 9 eaons of he classcal IFT algohm an sep esponse wh he nal paameesoe 56

59 We see ha he classcal IFT ceon els a smalle maxmm oveshoo an selng me. Usng a me weghng n he cos fncon, we wll now show how we can ge bee esls. onseng he cos fncon an applng a me weghng J N [ w ] w, w 5, [,.5] [.5,].5,.5 as shown n he followng fge Fg. 4. we ge he followng sep esponse: Fg. 4. ha els a mch bee behavo wh espec o he classcal IFT ceon an was sng onl 4 eaons. 57

60 58 4. The BFGS meho fo he seach econ The max R n he paamee pae law eemnes he pae econ an s heefoe ccal fo he pefomance of he algohm. We have seen ha a goo choce s o le R be an appoxmaon of he Hessan. Especall f ~ s small, he Gass-Newon econ T T N N R λ s a esable chose an he naal appoxmaon s: T T N es es es es N R λ Anohe goo chose s he Boen-Fleche-Golfab-ShannoBFGS meho, one of he qas-newon mehos. One of he mes of he qas-newon mehos s ha he goo esmaon of Hessan max s gven fom he gaens of he cos fncon J an he esgn paamees. BFGS meho s well known as a goo opmzaon meho [Hamamoo e al, 3]. The enewal law o esmae he Hessan base on BFGS meho s gven as follows. k k T k k T k k k k T k T k k k k s B s B s s B s z z z B B whee ' ' k k k k k k k k k k J J J J z s B B The spescp enoes he k-h eaon. The nal max vale of B s an aba posve efne max. Usall B s chosen o be an en max. The followng facs ae well known abo he BFGS meho: a f B k s smmec hen B k s smmec; b f B k s posve efne an z k T s k >, hen B k s posve efne. When z k T s k > s no sasfe, le B k B k >. Howeve sch a case selom occs. Noce ha he BFGS meho ses he same aa as he Gass-Newon meho. Ths he IFT paamee pang law becomes: J B γ.

61 4.3 The Hamamoo-Fka-Sge IFT Appoach We conse he wo.o.f. conol ssem epce n he followng fge. Fg. 4.3 lose loop ssem wh wo-egee-of-feeom conolle Snce an wo.o.f. conol ssem can be ansfome n o hs confgaon [Sge & Yoshkawa, 986], hee s no loss of geneal hee. In fges, P s he SISO plan, an K an F enoe he conolles o be esgne. The scala, an ae he plan np, s op an he efeence, especvel. The sgnal enoes an exenal es sgnal whch s se fo he esmaon of he pefomance of he close loop ssem an we assme ha we can choose abal. Noe ha all sbssems ae known excep fo he plan P, an all he sgnals ae obsevable. As we have seen befoe, he ssem T s he ese close loop moel whch s gven n avance, an T s he ese ajeco whch shol ack. One sngshng feae of he conol sce shown n he fge s as follows. Le T enoe he close loop ansfe fncon fom o, hen s known ha he so-calle cononal feeback pope hols, ha s, T T K s sasfe wheneve F T P hols. In ohe wos, K oes no pla an ole fo ackng pope n he nomnal case, an he man ole of K s o sppess he effec of sbances an plan nceanes [Sge & Yoshkawa, 986]. Whle appaenl, he ole of F s o specf he ackng pope. The conolles K an F ae sppose o be nqel eemne b he esgn paamees a an b especvel. We se he smbols K a an F b n oe o ncae he epenence on he paamees, explcl. Fo I/O sgnals, we se he smla smbols sch as a an b. Fom he above obsevaon, wol be naal o ne K a an F b o as o acheve low sensv an ese ackng pope, especvel. In he followng we assme ha one sablzng conolle pa K a, F b s gven n avance. An each expemen s pefome n he fne me neval [, f ]. 59

62 The saeg popose b Hamamoo, Fka an Sge HFS-meho s he sepaee nng of he feeback an feefowa conolles. In pacla, as fo he feeback conolle, he concenae on nng o acheve low sensv nsea of ackng pope, whle he feefowa conolle s ne fom he vewpon f comman ackng pope. - Feeback conolle nng In hs secon we concenae on he nng of he feeback conolle K a. In oe o acheve low sensv, hs new meho o mnmze he weghe sensv fncon WsSs n a cean case, whee Ws s he gven weghng ansfe fncon an s he sensv fncon. S s P s K s To achve hs goal he HFS meho pefoms he followng wo expemens A,B. Expemen A Se f an n fge 4.3. Then he conolle ssem s shown as he fge below. Fg. 4.4 Injec he es sgnal whch s calclae fom s W s n s whee n s a val whe sgnal whch has zeo mean an an appopae covaance. Then, le a an a be he coesponng I/O sgnals of he plan. Fo hese I/O sgnals, we solve he followng poblem. 6

63 FB onolle Desgn Poblem Fn he paamee a * whch mnmzes wh J x fb a a λa a x, x f x T τ x τ τ hogh eaon of expemens, whee λ a s a posve consan weghng scala. Noe ha he ansfe fncon fom o s eqal o he sensv fncon S. Theefoe, f f an λ a, he above cos s eqal o he sqae of he H nom of WS becase of he wheness of he val sgnal n. Now we appl he HFS meho, ha eves fom he sana IFT algohm base on Hjalmasson an Bkelan 998 an Hamamoo an Sge 999. The paamee a s pae b γ B ' a a a fb a ' whee enoes he vale of a a he -h enewal, an J s he gaen a of J a wh espec o a. As fo he sana IFT meho, he max B a gves he pae econ an can be fo example a Gass-Newon appoxmaon of he Hessan of J o a max fon wh he BFGS meho. In o scsson an followng example he scala γ s consee as a fxe sep-sze. ' Fs, we calclae J fom he I/O aa of expemens who an moel of P. The j-h en of J fb a ' J fb s gven b ', λ, ' ' fb, j a j a a a j a a whee ' a enoes he evave wh espec o a an he sbscp j means he j-h elemen of he veco. ' In oe o calclae J, we pefom he followng expemen. fb a J fb a Expemen B Se an n fg. 4.3 an njec he sgnal f a whch s obane b Expemen A, an le f a an f a be he coesponng I/O sgnals of he plan. 6

64 6 Snce PK a a a a a K hol fom Expemen A, he evave wh espec o a gves s ' ' ' ' ' a a j a a a j a a a j a a j a j f K PK P K PK PK P K PK PK ' ' ' ' ' a a j a a a j a j a a a j a j f K PK P K K K whee we se he elaons a a a PK P f a a a PK f fom Expemen B. Theefoe we can calclae ' a fb J fom he aa,,, ' ' a a a a hogh Expemens A an B. The pocee of feeback conolle nng s soppe when, fo a gven scala > ε n avance, ε < k a fb k a fb J J an we ega k a as he sb-opmal paamees * a.

65 - Feefowa conolle nng k Now we fx he feeback conolle as K K, an ne he feefowa conolle F b. The objecve hee s o make ack moe accael fo he gven efeence. Fo hs ppose we pefom he followng expemen. a Expemen Injec o he wo.o.f. conol ssem n fg. 4.3 wh F b an K, an le b an b be he coesponng I/O sgnals of he plan. The ssem s shown n he fge below. Fg. 4.5 Fo hs ssem we solve he followng poblem. FF onolle Desgn Poblem Assme he conolles K an F whch sablze he ssem shown n fg. b * 4.3 ae gven. Fn he paamee b whch mnmzes he cos fncon J ff b a λb b hogh eaon of expemens, whee λ b enoes he posve consan weghng scala. The IFT pocee s almos he same as n he case of FB conolle nng b eplacng J fb J ff, K F an a b. The j-h en of he gaen of he cos fncon s gven b: J ', λ, ' ' ff, j b j b b b j b b To calclae we nee he followng expemen. 63

66 Expemen D Se an n fg. 4.3 an njec he sgnal f, an le b an b be he coesponng I/O sgnals of he plan. Fom Expemens an D, he followng elaons hol: PK P PK K b F b T PK PK P b b Whle, fom he evave wh espec o b, we have: ' ' ' j b F j F j b PK ' ' P ' j b F j Fj b PK ' ' Theefoe, he aa,,, Expemens an D, an we can calclae b ae obane hogh b b b ' J ff fom hese aa. A SIMULATION EXAMPLE Le s noce a smple moel of a obo jon expesse b he Laplace fncon: G whee J s he jon momen of nea. s J s The example s aken fo [Scalamogna, ]. In ha wok, he ppose was o mpove he pefomance of he conolle ssem b ang o he ssem a sable exenal conol sgnal sng Ieave Leanng onol IL. We wll examne he same example n ems of IFT. 64

67 65 Fom G c s we oban Gz ha s he scee-me veson of G c s nclng he Zeo Oe Hol ZOH: z z G z z J T z G gan s whee T s s he samplng me. The jon s conolle b a PD feeback conolle Kz an b a fee-fowa conolle F f z. z k T k k z T k T k z z T k k z G s p s p s p s p hence z F z F z K zeo gan an z z F z z T J z F gan f s f wh J.94 Ns k p.7 k.4 T s. s The followng fge shows he scheme we ae conseng noce he sbance ae. Fg. 4.6 We choose he followng paamezaons of he wo conolles: z z z K z z z z F f

68 wh he nal paamees: The efeence ajeco chosen n he expemen s sn π. The nal behavo s shown n he pces below. Fg. 4.7 Fg. 4.8 We wsh o mpove he pefomance of he conolle ssem b sng he HFS- IFT appoach. 66

69 We pefom 3 eaon of he IFT algohm o ne he feeback conolle K wh a Gass-Newon appoxmaon of he Hessan fo he max B, a sep sze of. an λ a λ b. The fnal paamees ae: Fg. 4.9 an he coesponng ackng eo: B 3 eaon we pass fom an nal cos -5 J fb J fb.439 o a fnal cos The followng fge shows he anson of he cos J fb b eaons. Fg. 4. Fom hs fge we can see ha he cos J fb eceases b he meho popose b Hamamoo, Fka an Sge. 67

70 The followng fge shows he op esponse coesponng o he es sgnal. Fg. 4. Fll: esponse afe feeback nng Doe: esponse wh he nal conolle Ths fge shows ha lowe sensv s acheve va IFT. The gan an phase plos of sensv ae shown n he fge below. The obane conolle acheves a sasfaco pope n he low-feqenc oman. These fges show ha he popose IFT meho woks well n oe o acheve low sensv. Fg. 4. Sensv fncon Fll: fnal conolle afe feeback nng Doe: nal conolle 68

71 We now fx he feeback conolle as fon afe he pevos 3 eaons an we pefom eaons wh he ppose o ne he feefowa conolle F f. B en eaons, we pass fom an nal cos -3 J ff J ff.4 o a fnal cos The followng fge shows he anson of he cos J ff as a fncon of eaon nmbe. Fg. 4.3 The followng fge shows he ackng eo afe he nng of he feefowa conolle F f. Fg. 4.4 Tackng eo -Fll: afe feefowa nng - Doe: afe feeback nng - Dashe: nal 69

72 The fnal paamees ae: Smmazng, he HFS-meho se sepaae nng of he feeback an feefowa conolles. As fo he feeback conolle, he pon s focse on nng n oe o acheve low sensv nsea of ackng pope, whle he feefowa conolle have been ne fom he vewpon of enhancemen of ackng pefomance. The expemenal esls show he effecveness of he popose IFT meho. 7

73 5. Applcaons on an Insal Robo 5. The ABB nsal obo Ib- bef escpon The obo se n he expemens s an ABB Ib- nsal obo. The obo has seven lnks whch ae connece b sx jons, as shown n he followng fge. Fg. 5. Fg. 5. I s bl p b wo bg ams an a ws. Jon axs B n he fge s se o move he lowe am back an foh, wheeas jon 3 A moves he ppe am p an own. Jon 4 D s se o n he ws n an jon 5 E bens he ws n aon s cene. The sxh jon F s se o n he obo en effeco, whch s mone on he p of he ws he en effeco s no shown n he fge. Fnall, jon ns he ene obo aon s base. The obo ssem has ffeen bl-n conolles, one fo he conol of each jon angle. These conolles ae cascae PID conolles. Fg

74 5. The expemenal plafom The expemenal plafom consss of: - econfge Ib- obo ssem obo an conol cabne; - VME base boa compe ssem age ssem; - Hos compe ssem conssng of Sn woksaons hos ssem; - Ehene connecon beween hos an age. The Ib- s conolle fom VME-base embee compes. Sn woksaons ae se fo sofwae evelopmen an conol engneeng, as well as fo obo opeao neacon. PU-boa Powe P Fg. 5.4 The above fge shows he Ib- pa of he laboao. Sgnals fom nenal sensos of he obo o he VME ssem go va he senso neface o he DSP boa connece o he VME bs. The mase compe n he VME compe s base on Powe P pocesso. Spevson an safe fncons ae mplemene on a M683 boa, well sepaae fom he es of he ssem o peven amage of he obo. Dgal Sgnal PocessosDSP ae se fo low-level conol an fleng of sensos sgnals. Sensos eqng ve hgh aa banwhs ae connece ecl o he DSP boas. An aonal DSP boa belongs o he foce-oqe senso. 7

75 5.3 The Malab-obo connecon B he Sn woksaons on hos-level s possble o efne an we he pogams o conol he obo o sen efeences o be acke. Tha can be one nse he Malab envonmen. A Malab pogam calle Exc_hanle excaon hanle s avalable fo smple excaon expemens on he obo. Ths pogam can be se o efne veloc an poson efeences o he obo sevos. The nps can be seps, amps, snsos, nose an ohe aba sgnals fom he Malab wokspace. Fg. 5.5 A lo of sgnals can be ecoe ng he excaon. These ncle np oqes, poson measemens, ffeenae poson veloc, an foce an oqe measemens fom he foce senso. The ecoe sgnals can be expoe o he Malab wokspace fo plong an aa pocessng. Fg

76 5.4 The Smlnk conol neface A Smlnk conol neface s avalable o hanle he connecon beween he obo an he Malab envonmen. The fge below epesens one obo jon. Fg. 5.7 Wh hs neface, s possble o connec an log sgnals wh he Excaon hanle Exc_hanle an also o bl o own conolle ha we wll connec o he oqe_ef oqe efeence sgnal. Afe we have bl o conolle, he Smlnk scheme can be aomacall anslae no coe an ownloae ecl no he obo conol n. To be able o se o own conolle fo jon k we nee o acvae he new conolles fo he obo. Ths opeaon can be one sng he malab pogam acvae_smjk. To eacvae he ognal conolles fo he jon we se acvae_smj. Wh he malab pogams RegOff an RegOn we can emove an ensall, especvel, he ognal conolle fo he specfe -h jon. 74

77 5.5 Applcaon of IFT o a obo jon moel Befoe we come o he eal applcaon we wll s some esls of he IFT algohm fo nng a conolle fo a moel of he base jon jon fo he obo. The jon moel sce we conse s he followng: G s s a s s s ωξ s ω ω ξ s ω The vales se n he smlaons ae: a 7 ω 5 ω 7 ξ.4 ξ.45 Noce he pesence of an negao n he pocess moel. We conol he ssem wh a sana PID conol sce: - PID Robo jon Fg. 5.8 G s k p T s T s Fo he ealzaon we choose: wh τ.. G s k p T s τ s T s hoce of he nal PID paamees - AMIGO nng les We choose o sa wh nal paamees Kp, T an T fon sng he AMIGO consevave nng les [Åsöm & Häggln, 3]. 75

78 The AMIGO les, n he case when an negao s pesen n he pocess, can be esme n he followng wa. We make an open loop sep esponse expemen n whch we mease he qanes K v an L of he fge, ha ae he slope an he nesecon wh he me axs of he sagh lne, especvel. K V L Fg. 5.9 Open loop sep esponse of he ssem Gs Then he AMIGO les gve he followng qanes of he PID paamees: k T T p.45 K L 8L.5L In o specfc case, we mease he qanes: V K V L. So ha we ge he nal PID paamees fo he obo jon conolle: k T T p

79 These paamees el he followng nal sep esponse: Fg. 5. Inal sep esponse of he obo jon moel Now we appl he sana IFT algohm wh he am of mpovng he obo jon esponse. We cone he followng ese op esponse o he efeence sgnal chosen as a sep:. T. s 3s Fg. 5. Jon - Dese op 77

80 The applcaon of he classcal IFT ceon sng he cos fncon an conseng: J N N - a Gass-Newon appoxmaon of he Hessan fo he max R - a sep sze γ. els, afe especvel,, 3 an 4 eaons, he esponses: Fg. 5. Fll lne: fnal Jon sep esponse afe eaons Dashe: nal esponse - Doe: ese op Fg. 5.3 Fll lne: fnal Jon sep esponse afe eaons Dashe: nal esponse - Doe: ese op 78

81 Fg. 5.4 Fll lne: fnal Jon sep esponse afe 3 eaons Dashe: nal esponse - Doe: ese op Fg. 5.5 Fll lne: fnal Jon sep esponse afe 4 eaons Dashe: nal esponse - Doe: ese op 79

82 We can plo he gaph of he cos fncon J espec o he IFT eaons an smmaze he vales n he followng able. Fg. 5.6 eon as fncon of he nmbe of IFT eaons eon vales J nal.8744 J.56 J.98 J 3.6 J 4fnal.958 Fg. 5.7 The fnal vales of he PID paamees ae: k T T p

83 We can follow he paamee pae ng he IFT algohm lookng a he followng gaphs: Fg. 5.8 Fg. 5.9 Fg. 5. 8

84 5.6 APPLIATION OF IFT ON THE REAL ROBOT PROESS 5.6. THE JOINT EXPERIMENT We mplemen a PID conolle n he Smlnk obo neface as he followng he sana conolle s eacvae an he conol sgnal s ecl on he oqe. Fg. 5. The ppose s o mpove he esponse of he obo jon. The efeence ses s ef 4 a s a, wh a.5, n sep. 4 We sa wh he nal paamees: K T T p.4. We se he ceon N J an N - ese op as he efeence sgnal; - a Gass-Newon appoxmaon of he Hessan fo he max R ; - a sep sze γ.. We pefome hee ffeen IFT schemes: sana no weghs, mofe wh oble weghng an mofe wh vaable mask. The esls ae shown n he followng secons. 8

85 Sana IFT no weghs Fg. 5. Fll: nal esponse - Doe: ese op Fg. 5.3 Fll: fnal esponse afe eaons Dashe: nal Doe: ese We can noce how he selng me s no ve sasfaco. 83

86 In he followng pces he paamee paes ae shown. Fg. 5.4 Fg. 5.5 Fg

87 Mofe IFT fxe weghng We se now a mofe ceon sng me weghng as shown n he fs pce. wegh wegh 5.5 s wegh 3.75 s wegh 5 Fg. 5.7 Fll: nal esponse - Doe: ese op Fg. 5.8 Fll: fnal esponse afe 8 eaons Dashe: nal Doe: ese We can noce how hs echnqe els o a bee esponse n ems of selng me. Howeve we noce oscllaons ng he oveshoo an he esl fom eaon nmbe 9 became ccal sabl poblem. 85

88 The paamee paes ae shown n he pces below. Fg. 5.9 Fg. 5.3 Fg

89 Mofe IFT Weghng wh a vaable mask A mask wh nal leng of.5 secons s se as shown n he fge below an a eve eaon s ecease b.5 secons. wegh wegh wegh wegh Fg. 5.3 Fll: nal esponse - Doe: ese op Fg Fll: fnal esponse afe eaons Dashe: nal Doe: ese 87

90 The paamee paes ae shown n he pce below. Fg Fg Fg

91 5.6. THE FLEXIBLE BEAM EXPERIMENT The ppose s o esgn a conolle fo he beam eflecon an o mpove he behavo of he ssem sng he IFT algohm. We se p he obo wh a flexble beam mone on he jon 6 of he obo see fge below. Fg The flexble beam We have se a foce/oqe senso JR3 fo measng he beam eflecon oqe appoxmael popoonal o he eflecon. The beam was also eqppe wh a san gage whch also col be se fo esmang he beam eflecon b n hese expemens we onl se oqe measemens. The ssems s songl elae wh he wo-mass-pocess seen n he pevos chape, whee we col conse he fs mass as he obo mass an he secon mass as he beam mass. The pocess col be compae wh he pce below, whee m >>m. Fg The wo-mass-pocess 89

92 Ienfcaon Expemen The fs sep s o enf a SIMO moel Sngle Inp Mlple Op fom he poson efeencesgnal pos ef o he obo poson an beam efleconsgnal m whch s o be se fo conol of he beam eflecon. In hese expemens we have he sana poson conolle acvae. Sep esponse expemen Fg Fg. 5.4 We see ha he poson conol of he obo jon s sasfaco b hee ae lage pool ampe oscllaons fo he flexble beam ~6 Hz. 9

93 Feqenc esponse expemen To esmae a goo moel fo he conol we se a Pseo Ranom Bna Seqence PRBS as excaon sgnal. Fg. 5.4 Response of he jon poson wh a PRBS as efeence Fg. 5.4 Response of he flexble beam evaon sgnal m wh a PRBS as efeence 9

94 Usng a sb-space esmaon meho N4SID fom Ssem Ienfcaon Toolbox Malab, we ge goo esls fo a sae space moel of oe 6: loss fncon.394, FPE.445. Fg Feqenc esponse of he 6 oe moel fon wh he N4SID meho pos_ef o m We o ece o oe 4 o cape essenal namcs b, wh he ognal aa seqence we have a poblem of machng he coec fs esonanc feqenc an we have a lowe peak fo he gan. We ge he followng vales: loss fncon.93784, FPE.44. Fg Feqenc esponse of he 4 oe moel fll lne fon wh he N4SID meho pos ef o m an of he 6 oe moel oe 9

95 Fnall, sng a low pass fle on he sgnals befoe ong he enfcaon n oe o mach he fs esonance feqenc, we ge, mplemenng agan he N4SID meho, a sae space moel of oe 4 wh loss fncon.94e-6 an FPE.436e-6 an he followng feqenc esponse. Fg Feqenc esponse of he 4 oe moel fll lne fon wh he N4SID meho pos ef o m sng flee aa Doe: pevos 6 oe moel Dashe: pevos 4 oe moel no pe-fleng Fg Flee veson of he aa se se fo he enfcaon of he foh-oe moel wh pefleng 93

96 The fnal scee sae-space moel enfe oe 4, pefleng an chosen fo he conol mplemenaon s: x s Ax B x D wh he samplng me s.5 an he maces: A B D The poles of he enfe moel ae: p p p p j j.547 j j.736 Fg

97 ONTROL DESIGN We choose a sae feeback conol sce wh he ppose o amp he flexble beam eflecon m an sll ge a goo fas sep esponse. Pole placemen Fg Fo he close-loop ssem we choose o keep he spee an he ampng of he wo poles coesponng o he poson an move he wo pool ampe poles elae o he flexble beam keepng he same egenfeqenc b nceasng he ampng. Fom he enfe scee me moel he poles wee ansfee o he coesponng connos me poles wee we have an eas nepeaon of he ampng. z a z a s ωζs ω ps Poles ae mappe accong o z e p connos me pole, z scee me pole [Åsöm & Wenmak]. In o case we ncease he ampng of he complex poles fom appox. o.74 We make a pole placemen esgn wh he new scee me poles. p p p p j.6 j.6 j.3 j.3 Fg

Optimal Control Strategies for Speed Control of Permanent-Magnet Synchronous Motor Drives

Optimal Control Strategies for Speed Control of Permanent-Magnet Synchronous Motor Drives Wol Acaemy of Scence, Engneeng an echnology 44 8 Opmal Conol Saeges fo Spee Conol of Pemanen-Magne Synchonos Moo Dves Roozbeh Molav, an Davoo A. Khab Absac he pemanen magne synchonos moo (PMSM) s vey sefl

More information

Chapter Finite Difference Method for Ordinary Differential Equations

Chapter Finite Difference Method for Ordinary Differential Equations Chape 8.7 Fne Dffeence Mehod fo Odnay Dffeenal Eqaons Afe eadng hs chape, yo shold be able o. Undesand wha he fne dffeence mehod s and how o se o solve poblems. Wha s he fne dffeence mehod? The fne dffeence

More information

I-POLYA PROCESS AND APPLICATIONS Leda D. Minkova

I-POLYA PROCESS AND APPLICATIONS Leda D. Minkova The XIII Inenaonal Confeence Appled Sochasc Models and Daa Analyss (ASMDA-009) Jne 30-Jly 3, 009, Vlns, LITHUANIA ISBN 978-9955-8-463-5 L Sakalaskas, C Skadas and E K Zavadskas (Eds): ASMDA-009 Seleced

More information

5-1. We apply Newton s second law (specifically, Eq. 5-2). F = ma = ma sin 20.0 = 1.0 kg 2.00 m/s sin 20.0 = 0.684N. ( ) ( )

5-1. We apply Newton s second law (specifically, Eq. 5-2). F = ma = ma sin 20.0 = 1.0 kg 2.00 m/s sin 20.0 = 0.684N. ( ) ( ) 5-1. We apply Newon s second law (specfcally, Eq. 5-). (a) We fnd he componen of he foce s ( ) ( ) F = ma = ma cos 0.0 = 1.00kg.00m/s cos 0.0 = 1.88N. (b) The y componen of he foce s ( ) ( ) F = ma = ma

More information

1 Constant Real Rate C 1

1 Constant Real Rate C 1 Consan Real Rae. Real Rae of Inees Suppose you ae equally happy wh uns of he consumpon good oday o 5 uns of he consumpon good n peod s me. C 5 Tha means you ll be pepaed o gve up uns oday n eun fo 5 uns

More information

Physics 201 Lecture 15

Physics 201 Lecture 15 Phscs 0 Lecue 5 l Goals Lecue 5 v Elo consevaon of oenu n D & D v Inouce oenu an Iulse Coens on oenu Consevaon l oe geneal han consevaon of echancal eneg l oenu Consevaon occus n sses wh no ne eenal foces

More information

Suppose we have observed values t 1, t 2, t n of a random variable T.

Suppose we have observed values t 1, t 2, t n of a random variable T. Sppose we have obseved vales, 2, of a adom vaable T. The dsbo of T s ow o belog o a cea ype (e.g., expoeal, omal, ec.) b he veco θ ( θ, θ2, θp ) of ow paamees assocaed wh s ow (whee p s he mbe of ow paamees).

More information

CHAPTER 10: LINEAR DISCRIMINATION

CHAPTER 10: LINEAR DISCRIMINATION HAPER : LINEAR DISRIMINAION Dscmnan-based lassfcaon 3 In classfcaon h K classes ( k ) We defned dsmnan funcon g () = K hen gven an es eample e chose (pedced) s class label as f g () as he mamum among g

More information

2 shear strain / L for small angle

2 shear strain / L for small angle Sac quaons F F M al Sess omal sess foce coss-seconal aea eage Shea Sess shea sess shea foce coss-seconal aea llowable Sess Faco of Safe F. S San falue Shea San falue san change n lengh ognal lengh Hooke

More information

CptS 570 Machine Learning School of EECS Washington State University. CptS Machine Learning 1

CptS 570 Machine Learning School of EECS Washington State University. CptS Machine Learning 1 ps 57 Machne Leann School of EES Washnon Sae Unves ps 57 - Machne Leann Assume nsances of classes ae lneal sepaable Esmae paamees of lnea dscmnan If ( - -) > hen + Else - ps 57 - Machne Leann lassfcaon

More information

Rotations.

Rotations. oons j.lbb@phscs.o.c.uk To s summ Fmes of efeence Invnce une nsfomons oon of wve funcon: -funcons Eule s ngles Emple: e e - - Angul momenum s oon geneo Genec nslons n Noehe s heoem Fmes of efeence Conse

More information

Chapter 3: Vectors and Two-Dimensional Motion

Chapter 3: Vectors and Two-Dimensional Motion Chape 3: Vecos and Two-Dmensonal Moon Vecos: magnude and decon Negae o a eco: eese s decon Mulplng o ddng a eco b a scala Vecos n he same decon (eaed lke numbes) Geneal Veco Addon: Tangle mehod o addon

More information

Backcalculation Analysis of Pavement-layer Moduli Using Pattern Search Algorithms

Backcalculation Analysis of Pavement-layer Moduli Using Pattern Search Algorithms Bakallaon Analyss of Pavemen-laye Modl Usng Paen Seah Algohms Poje Repo fo ENCE 74 Feqan Lo May 7 005 Bakallaon Analyss of Pavemen-laye Modl Usng Paen Seah Algohms. Inodon. Ovevew of he Poje 3. Objeve

More information

Name of the Student:

Name of the Student: Engneeng Mahemacs 05 SUBJEC NAME : Pobably & Random Pocess SUBJEC CODE : MA645 MAERIAL NAME : Fomula Maeal MAERIAL CODE : JM08AM007 REGULAION : R03 UPDAED ON : Febuay 05 (Scan he above QR code fo he dec

More information

Integral Control via Bias Estimation

Integral Control via Bias Estimation 1 Integal Contol via Bias stimation Consie the sstem ẋ = A + B +, R n, R p, R m = C +, R q whee is an nknown constant vecto. It is possible to view as a step istbance: (t) = 0 1(t). (If in fact (t) vaies

More information

TRANSIENTS. Lecture 5 ELEC-E8409 High Voltage Engineering

TRANSIENTS. Lecture 5 ELEC-E8409 High Voltage Engineering TRANSIENTS Lece 5 ELECE8409 Hgh Volage Engneeng TRANSIENT VOLTAGES A ansen even s a sholved oscllaon (sgnfcanly fase han opeang feqency) n a sysem cased by a sdden change of volage, cen o load. Tansen

More information

( ) ( ) Weibull Distribution: k ti. u u. Suppose t 1, t 2, t n are times to failure of a group of n mechanisms. The likelihood function is

( ) ( ) Weibull Distribution: k ti. u u. Suppose t 1, t 2, t n are times to failure of a group of n mechanisms. The likelihood function is Webll Dsbo: Des Bce Dep of Mechacal & Idsal Egeeg The Uvesy of Iowa pdf: f () exp Sppose, 2, ae mes o fale of a gop of mechasms. The lelhood fco s L ( ;, ) exp exp MLE: Webll 3//2002 page MLE: Webll 3//2002

More information

L-1. Intertemporal Trade in a Two- Period Model

L-1. Intertemporal Trade in a Two- Period Model L-. neempoal Tade n a Two- Peod Model Jaek Hník www.jaom-hnk.wbs.z Wha o Shold Alead now en aon def... s a esl of expos fallng sho of mpos. s a esl of savngs fallng sho of nvesmens. S A B NX G B B M X

More information

EE 410/510: Electromechanical Systems Chapter 3

EE 410/510: Electromechanical Systems Chapter 3 EE 4/5: Eleomehnl Syem hpe 3 hpe 3. Inoon o Powe Eleon Moelng n Applon of Op. Amp. Powe Amplfe Powe onvee Powe Amp n Anlog onolle Swhng onvee Boo onvee onvee Flyb n Fow onvee eonn n Swhng onvee 5// All

More information

Modern Energy Functional for Nuclei and Nuclear Matter. By: Alberto Hinojosa, Texas A&M University REU Cyclotron 2008 Mentor: Dr.

Modern Energy Functional for Nuclei and Nuclear Matter. By: Alberto Hinojosa, Texas A&M University REU Cyclotron 2008 Mentor: Dr. Moden Enegy Funconal fo Nucle and Nuclea Mae By: lbeo noosa Teas &M Unvesy REU Cycloon 008 Meno: D. Shalom Shlomo Oulne. Inoducon.. The many-body poblem and he aee-fock mehod. 3. Skyme neacon. 4. aee-fock

More information

BISTATIC COHERENT MIMO CLUTTER RANK ANALYSIS

BISTATIC COHERENT MIMO CLUTTER RANK ANALYSIS 3 Euopean Sgnal Pocessng Confeence (EUSIPCO BISAIC COHEEN MIMO CLUE ANK ANALYSIS Ksne Bell, * Joel Johnson, Chsophe Bae, Gaeme Smh, an Mualha angaswam * Meon, Inc, 88 Lba S, Sue 600, eson, Vgna 090, USA

More information

Lecture 5. Plane Wave Reflection and Transmission

Lecture 5. Plane Wave Reflection and Transmission Lecue 5 Plane Wave Reflecon and Tansmsson Incden wave: 1z E ( z) xˆ E (0) e 1 H ( z) yˆ E (0) e 1 Nomal Incdence (Revew) z 1 (,, ) E H S y (,, ) 1 1 1 Refleced wave: 1z E ( z) xˆ E E (0) e S H 1 1z H (

More information

New Stability Condition of T-S Fuzzy Systems and Design of Robust Flight Control Principle

New Stability Condition of T-S Fuzzy Systems and Design of Robust Flight Control Principle 96 JOURNAL O ELECRONIC SCIENCE AND ECHNOLOGY, VOL., NO., MARCH 3 New Sably Conon of -S uzzy Sysems an Desgn of Robus lgh Conol Pncple Chun-Nng Yang, Ya-Zhou Yue, an Hu L Absac Unlke he pevous eseach woks

More information

Field due to a collection of N discrete point charges: r is in the direction from

Field due to a collection of N discrete point charges: r is in the direction from Physcs 46 Fomula Shee Exam Coulomb s Law qq Felec = k ˆ (Fo example, f F s he elecc foce ha q exes on q, hen ˆ s a un veco n he decon fom q o q.) Elecc Feld elaed o he elecc foce by: Felec = qe (elecc

More information

The sound field of moving sources

The sound field of moving sources Nose Engneeng / Aoss -- ong Soes The son el o mong soes ong pon soes The pesse el geneae by pon soe o geneal me an The pess T poson I he soe s onenae a he sngle mong pon, soe may I he soe s I be wen as

More information

by Lauren DeDieu Advisor: George Chen

by Lauren DeDieu Advisor: George Chen b Laren DeDe Advsor: George Chen Are one of he mos powerfl mehods o nmercall solve me dependen paral dfferenal eqaons PDE wh some knd of snglar shock waves & blow-p problems. Fed nmber of mesh pons Moves

More information

Chapter Finite Difference Method for Ordinary Differential Equations

Chapter Finite Difference Method for Ordinary Differential Equations Chape 8.7 Finie Diffeence Mehod fo Odinay Diffeenial Eqaions Afe eading his chape, yo shold be able o. Undesand wha he finie diffeence mehod is and how o se i o solve poblems. Wha is he finie diffeence

More information

ScienceDirect. Behavior of Integral Curves of the Quasilinear Second Order Differential Equations. Alma Omerspahic *

ScienceDirect. Behavior of Integral Curves of the Quasilinear Second Order Differential Equations. Alma Omerspahic * Avalable onlne a wwwscencedeccom ScenceDec oceda Engneeng 69 4 85 86 4h DAAAM Inenaonal Smposum on Inellgen Manufacung and Auomaon Behavo of Inegal Cuves of he uaslnea Second Ode Dffeenal Equaons Alma

More information

3. A Review of Some Existing AW (BT, CT) Algorithms

3. A Review of Some Existing AW (BT, CT) Algorithms 3. A Revew of Some Exstng AW (BT, CT) Algothms In ths secton, some typcal ant-wndp algothms wll be descbed. As the soltons fo bmpless and condtoned tansfe ae smla to those fo ant-wndp, the pesented algothms

More information

Variants of Pegasos. December 11, 2009

Variants of Pegasos. December 11, 2009 Inroducon Varans of Pegasos SooWoong Ryu bshboy@sanford.edu December, 009 Youngsoo Cho yc344@sanford.edu Developng a new SVM algorhm s ongong research opc. Among many exng SVM algorhms, we wll focus on

More information

( ) ( )) ' j, k. These restrictions in turn imply a corresponding set of sample moment conditions:

( ) ( )) ' j, k. These restrictions in turn imply a corresponding set of sample moment conditions: esng he Random Walk Hypohess If changes n a sees P ae uncoelaed, hen he followng escons hold: va + va ( cov, 0 k 0 whee P P. k hese escons n un mply a coespondng se of sample momen condons: g µ + µ (,,

More information

Determination of residual stresses and material properties by an energy-based method using artificial neural networks

Determination of residual stresses and material properties by an energy-based method using artificial neural networks 296 Poceedngs of he Esonan Academ of Scences, 2012, 61, 4, 296 305 Poceedngs of he Esonan Academ of Scences, 2012, 61, 4, 296 305 do: 10.3176/poc.2012.4.04 Avalable onlne a www.eap.ee/poceedngs Deemnaon

More information

Outline. GW approximation. Electrons in solids. The Green Function. Total energy---well solved Single particle excitation---under developing

Outline. GW approximation. Electrons in solids. The Green Function. Total energy---well solved Single particle excitation---under developing Peenaon fo Theoecal Condened Mae Phyc n TU Beln Geen-Funcon and GW appoxmaon Xnzheng L Theoy Depamen FHI May.8h 2005 Elecon n old Oulne Toal enegy---well olved Sngle pacle excaon---unde developng The Geen

More information

ajanuary't I11 F or,'.

ajanuary't I11 F or,'. ',f,". ; q - c. ^. L.+T,..LJ.\ ; - ~,.,.,.,,,E k }."...,'s Y l.+ : '. " = /.. :4.,Y., _.,,. "-.. - '// ' 7< s k," ;< - " fn 07 265.-.-,... - ma/ \/ e 3 p~~f v-acecu ean d a e.eng nee ng sn ~yoo y namcs

More information

, t 1. Transitions - this one was easy, but in general the hardest part is choosing the which variables are state and control variables

, t 1. Transitions - this one was easy, but in general the hardest part is choosing the which variables are state and control variables Opmal Conrol Why Use I - verss calcls of varaons, opmal conrol More generaly More convenen wh consrans (e.g., can p consrans on he dervaves More nsghs no problem (a leas more apparen han hrogh calcls of

More information

Continuous-time evolutionary stock and bond markets with time-dependent strategies

Continuous-time evolutionary stock and bond markets with time-dependent strategies Afcan Jounal of Busness Managemen Vol. 64 pp. 463-474 Febuay Avalable onlne a hp://www.acaemcjounals.og/ajbm DOI:.5897/AJBM.5 ISSN 993-833 Acaemc Jounals Full Lengh Reseach Pape Connuous-me evoluonay soc

More information

ON TOTAL TIME ON TEST TRANSFORM ORDER ABSTRACT

ON TOTAL TIME ON TEST TRANSFORM ORDER ABSTRACT V M Chacko E CONVE AND INCREASIN CONVE OAL IME ON ES RANSORM ORDER R&A # 4 9 Vol. Decembe ON OAL IME ON ES RANSORM ORDER V. M. Chacko Depame of Sascs S. homas Collee hss eala-68 Emal: chackovm@mal.com

More information

s = rθ Chapter 10: Rotation 10.1: What is physics?

s = rθ Chapter 10: Rotation 10.1: What is physics? Chape : oaon Angula poson, velocy, acceleaon Consan angula acceleaon Angula and lnea quanes oaonal knec enegy oaonal nea Toque Newon s nd law o oaon Wok and oaonal knec enegy.: Wha s physcs? In pevous

More information

Today - Lecture 13. Today s lecture continue with rotations, torque, Note that chapters 11, 12, 13 all involve rotations

Today - Lecture 13. Today s lecture continue with rotations, torque, Note that chapters 11, 12, 13 all involve rotations Today - Lecue 13 Today s lecue coninue wih oaions, oque, Noe ha chapes 11, 1, 13 all inole oaions slide 1 eiew Roaions Chapes 11 & 1 Viewed fom aboe (+z) Roaional, o angula elociy, gies angenial elociy

More information

Notes on Optimal Control

Notes on Optimal Control F.L. Lews Moncef-O Donnell Endowed Cha Head Conols & Sensos Gop Aomaon & Robocs Reseach Inse ARRI he Unvesy of eas a Alngon Noes on Opmal Conol Sppoed by : NSF - PAUL WERBOS ARO RANDY ZACHERY Dagna abe

More information

N 1. Time points are determined by the

N 1. Time points are determined by the upplemena Mehods Geneaon of scan sgnals In hs secon we descbe n deal how scan sgnals fo 3D scannng wee geneaed. can geneaon was done n hee seps: Fs, he dve sgnal fo he peo-focusng elemen was geneaed o

More information

ENGI 4430 Advanced Calculus for Engineering Faculty of Engineering and Applied Science Problem Set 9 Solutions [Theorems of Gauss and Stokes]

ENGI 4430 Advanced Calculus for Engineering Faculty of Engineering and Applied Science Problem Set 9 Solutions [Theorems of Gauss and Stokes] ENGI 44 Avance alculus fo Engineeing Faculy of Engineeing an Applie cience Poblem e 9 oluions [Theoems of Gauss an okes]. A fla aea A is boune by he iangle whose veices ae he poins P(,, ), Q(,, ) an R(,,

More information

CHAPTER 3 DETECTION TECHNIQUES FOR MIMO SYSTEMS

CHAPTER 3 DETECTION TECHNIQUES FOR MIMO SYSTEMS 4 CAPTER 3 DETECTION TECNIQUES FOR MIMO SYSTEMS 3. INTRODUCTION The man challenge n he paccal ealzaon of MIMO weless sysems les n he effcen mplemenaon of he deeco whch needs o sepaae he spaally mulplexed

More information

CONSISTENT EARTHQUAKE ACCELERATION AND DISPLACEMENT RECORDS

CONSISTENT EARTHQUAKE ACCELERATION AND DISPLACEMENT RECORDS APPENDX J CONSSTENT EARTHQUAKE ACCEERATON AND DSPACEMENT RECORDS Earhqake Acceleraons can be Measred. However, Srcres are Sbjeced o Earhqake Dsplacemens J. NTRODUCTON { XE "Acceleraon Records" }A he presen

More information

Using DP for hierarchical discretization of continuous attributes. Amit Goyal (31 st March 2008)

Using DP for hierarchical discretization of continuous attributes. Amit Goyal (31 st March 2008) Usng DP fo heachcal dscetzaton of contnos attbtes Amt Goyal 31 st Mach 2008 Refeence Chng-Cheng Shen and Yen-Lang Chen. A dynamc-pogammng algothm fo heachcal dscetzaton of contnos attbtes. In Eopean Jonal

More information

Reflection and Refraction

Reflection and Refraction Chape 1 Reflecon and Refacon We ae now neesed n eplong wha happens when a plane wave avelng n one medum encounes an neface (bounday) wh anohe medum. Undesandng hs phenomenon allows us o undesand hngs lke:

More information

L4:4. motion from the accelerometer. to recover the simple flutter. Later, we will work out how. readings L4:3

L4:4. motion from the accelerometer. to recover the simple flutter. Later, we will work out how. readings L4:3 elave moon L4:1 To appl Newon's laws we need measuemens made fom a 'fed,' neal efeence fame (unacceleaed, non-oang) n man applcaons, measuemens ae made moe smpl fom movng efeence fames We hen need a wa

More information

International Journal on Organic Electronics (IJOE) Vol.7, No.1, January E.L. Pankratov

International Journal on Organic Electronics (IJOE) Vol.7, No.1, January E.L. Pankratov ON OPMZAON OF MANUFAURNG OF A MO POR AMPFR O NRA NY OF MN H AOUN M-MAH NU R AN POROY OF MARA.. Panaov Nhn Novgoo ae Unves Gagan avene Nhn Novgoo 695 Rssa ABRA n hs pape we noce an appoach o ncease ens

More information

Comparative Study of Inventory Model for Duopolistic Market under Trade Credit Deepa H Kandpal *, Khimya S Tinani #

Comparative Study of Inventory Model for Duopolistic Market under Trade Credit Deepa H Kandpal *, Khimya S Tinani # Inenaonal Jounal of Saska an Mahemaka ISSN: 77-79 E-ISSN: 49-865 Volume 6 Issue pp -9 ompaave Suy of Invenoy Moel fo Duopolsc Make une ae e Deepa H Kanpal * Khmya S nan # Depamen of Sascs Faculy of Scence

More information

FIRMS IN THE TWO-PERIOD FRAMEWORK (CONTINUED)

FIRMS IN THE TWO-PERIOD FRAMEWORK (CONTINUED) FIRMS IN THE TWO-ERIO FRAMEWORK (CONTINUE) OCTOBER 26, 2 Model Sucue BASICS Tmelne of evens Sa of economc plannng hozon End of economc plannng hozon Noaon : capal used fo poducon n peod (decded upon n

More information

Observer Design for Nonlinear Systems using Linear Approximations

Observer Design for Nonlinear Systems using Linear Approximations Observer Desgn for Nonlnear Ssems sng Lnear Appromaons C. Navarro Hernandez, S.P. Banks and M. Aldeen Deparmen of Aomac Conrol and Ssems Engneerng, Unvers of Sheffeld, Mappn Sree, Sheffeld S 3JD. e-mal:

More information

J i-1 i. J i i+1. Numerical integration of the diffusion equation (I) Finite difference method. Spatial Discretization. Internal nodes.

J i-1 i. J i i+1. Numerical integration of the diffusion equation (I) Finite difference method. Spatial Discretization. Internal nodes. umercal negraon of he dffuson equaon (I) Fne dfference mehod. Spaal screaon. Inernal nodes. R L V For hermal conducon le s dscree he spaal doman no small fne spans, =,,: Balance of parcles for an nernal

More information

Simulation of Non-normal Autocorrelated Variables

Simulation of Non-normal Autocorrelated Variables Jounal of Moden Appled Sascal Mehods Volume 5 Issue Acle 5 --005 Smulaon of Non-nomal Auocoelaed Vaables HT Holgesson Jönöpng Inenaonal Busness School Sweden homasholgesson@bshse Follow hs and addonal

More information

ESS 265 Spring Quarter 2005 Kinetic Simulations

ESS 265 Spring Quarter 2005 Kinetic Simulations SS 65 Spng Quae 5 Knec Sulaon Lecue une 9 5 An aple of an lecoagnec Pacle Code A an eaple of a knec ulaon we wll ue a one denonal elecoagnec ulaon code called KMPO deeloped b Yohhau Oua and Hoh Mauoo.

More information

Stochastic Programming handling CVAR in objective and constraint

Stochastic Programming handling CVAR in objective and constraint Sochasc Programmng handlng CVAR n obecve and consran Leondas Sakalaskas VU Inse of Mahemacs and Informacs Lhana ICSP XIII Jly 8-2 23 Bergamo Ialy Olne Inrodcon Lagrangan & KKT condons Mone-Carlo samplng

More information

Lecture 18: Kinetics of Phase Growth in a Two-component System: general kinetics analysis based on the dilute-solution approximation

Lecture 18: Kinetics of Phase Growth in a Two-component System: general kinetics analysis based on the dilute-solution approximation Lecue 8: Kineics of Phase Gowh in a Two-componen Sysem: geneal kineics analysis based on he dilue-soluion appoximaion Today s opics: In he las Lecues, we leaned hee diffeen ways o descibe he diffusion

More information

THIS PAGE DECLASSIFIED IAW EO IRIS u blic Record. Key I fo mation. Ma n: AIR MATERIEL COMM ND. Adm ni trative Mar ings.

THIS PAGE DECLASSIFIED IAW EO IRIS u blic Record. Key I fo mation. Ma n: AIR MATERIEL COMM ND. Adm ni trative Mar ings. T H S PA G E D E CLA SSFED AW E O 2958 RS u blc Recod Key fo maon Ma n AR MATEREL COMM ND D cumen Type Call N u b e 03 V 7 Rcvd Rel 98 / 0 ndexe D 38 Eneed Dae RS l umbe 0 0 4 2 3 5 6 C D QC d Dac A cesson

More information

Fast Calibration for Robot Welding System with Laser Vision

Fast Calibration for Robot Welding System with Laser Vision Fas Calbaon fo Robo Weldng Ssem h Lase Vson Lu Su Mechancal & Eleccal Engneeng Nanchang Unves Nanchang, Chna Wang Guoong Mechancal Engneeng Souh Chna Unves of echnolog Guanghou, Chna Absac Camea calbaon

More information

ÖRNEK 1: THE LINEAR IMPULSE-MOMENTUM RELATION Calculate the linear momentum of a particle of mass m=10 kg which has a. kg m s

ÖRNEK 1: THE LINEAR IMPULSE-MOMENTUM RELATION Calculate the linear momentum of a particle of mass m=10 kg which has a. kg m s MÜHENDİSLİK MEKANİĞİ. HAFTA İMPULS- MMENTUM-ÇARPIŞMA Linea oenu of a paicle: The sybol L denoes he linea oenu and is defined as he ass ies he elociy of a paicle. L ÖRNEK : THE LINEAR IMPULSE-MMENTUM RELATIN

More information

Nanoparticles. Educts. Nucleus formation. Nucleus. Growth. Primary particle. Agglomeration Deagglomeration. Agglomerate

Nanoparticles. Educts. Nucleus formation. Nucleus. Growth. Primary particle. Agglomeration Deagglomeration. Agglomerate ucs Nucleus Nucleus omaon cal supesauaon Mng o eucs, empeaue, ec. Pmay pacle Gowh Inegaon o uson-lme pacle gowh Nanopacles Agglomeaon eagglomeaon Agglomeae Sablsaon o he nanopacles agans agglomeaon! anspo

More information

P a g e 5 1 of R e p o r t P B 4 / 0 9

P a g e 5 1 of R e p o r t P B 4 / 0 9 P a g e 5 1 of R e p o r t P B 4 / 0 9 J A R T a l s o c o n c l u d e d t h a t a l t h o u g h t h e i n t e n t o f N e l s o n s r e h a b i l i t a t i o n p l a n i s t o e n h a n c e c o n n e

More information

Data Flow Anomaly Analysis

Data Flow Anomaly Analysis Pof. D. Liggesmeye, 1 Contents Data flows an ata flow anomalies State machine fo ata flow anomaly analysis Example withot loops Example with loops Data Flow Anomaly Analysis Softwae Qality Assance Softwae

More information

Notes on Inductance and Circuit Transients Joe Wolfe, Physics UNSW. Circuits with R and C. τ = RC = time constant

Notes on Inductance and Circuit Transients Joe Wolfe, Physics UNSW. Circuits with R and C. τ = RC = time constant Nes n Inducance and cu Tansens Je Wlfe, Physcs UNSW cus wh and - Wha happens when yu clse he swch? (clse swch a 0) - uen flws ff capac, s d Acss capac: Acss ess: c d s d d ln + cns. 0, ln cns. ln ln ln

More information

Derivatives of Inverse Trig Functions

Derivatives of Inverse Trig Functions Derivaives of Inverse Trig Fncions Ne we will look a he erivaives of he inverse rig fncions. The formlas may look complicae, b I hink yo will fin ha hey are no oo har o se. Yo will js have o be carefl

More information

Part V: Velocity and Acceleration Analysis of Mechanisms

Part V: Velocity and Acceleration Analysis of Mechanisms Pat V: Velocty an Acceleaton Analyss of Mechansms Ths secton wll evew the most common an cuently pactce methos fo completng the knematcs analyss of mechansms; escbng moton though velocty an acceleaton.

More information

Real-Time Trajectory Generation and Tracking for Cooperative Control Systems

Real-Time Trajectory Generation and Tracking for Cooperative Control Systems Real-Tme Trajecor Generaon and Trackng for Cooperave Conrol Ssems Rchard Mrra Jason Hcke Calforna Inse of Technolog MURI Kckoff Meeng 14 Ma 2001 Olne I. Revew of prevos work n rajecor generaon and rackng

More information

( m is the length of columns of A ) spanned by the columns of A : . Select those columns of B that contain a pivot; say those are Bi

( m is the length of columns of A ) spanned by the columns of A : . Select those columns of B that contain a pivot; say those are Bi Assgmet /MATH 47/Wte Due: Thusday Jauay The poblems to solve ae umbeed [] to [] below Fst some explaatoy otes Fdg a bass of the colum-space of a max ad povg that the colum ak (dmeso of the colum space)

More information

( ) () we define the interaction representation by the unitary transformation () = ()

( ) () we define the interaction representation by the unitary transformation () = () Hgher Order Perurbaon Theory Mchael Fowler 3/7/6 The neracon Represenaon Recall ha n he frs par of hs course sequence, we dscussed he chrödnger and Hesenberg represenaons of quanum mechancs here n he chrödnger

More information

FAIPA_SAND: An Interior Point Algorithm for Simultaneous ANalysis and Design Optimization

FAIPA_SAND: An Interior Point Algorithm for Simultaneous ANalysis and Design Optimization FAIPA_AN: An Ineror Pon Algorhm for mlaneos ANalyss an esgn Opmzaon osé Hersos*, Palo Mappa* an onel llen** *COPPE / Feeral Unersy of Ro e anero, Mechancal Engneerng Program, Caa Posal 6853, 945 97 Ro

More information

Graduate Macroeconomics 2 Problem set 5. - Solutions

Graduate Macroeconomics 2 Problem set 5. - Solutions Graduae Macroeconomcs 2 Problem se. - Soluons Queson 1 To answer hs queson we need he frms frs order condons and he equaon ha deermnes he number of frms n equlbrum. The frms frs order condons are: F K

More information

PHYS 1443 Section 001 Lecture #4

PHYS 1443 Section 001 Lecture #4 PHYS 1443 Secon 001 Lecure #4 Monda, June 5, 006 Moon n Two Dmensons Moon under consan acceleraon Projecle Moon Mamum ranges and heghs Reerence Frames and relae moon Newon s Laws o Moon Force Newon s Law

More information

Sharif University of Technology - CEDRA By: Professor Ali Meghdari

Sharif University of Technology - CEDRA By: Professor Ali Meghdari Shaif Univesiy of echnology - CEDRA By: Pofesso Ali Meghai Pupose: o exen he Enegy appoach in eiving euaions of oion i.e. Lagange s Meho fo Mechanical Syses. opics: Genealize Cooinaes Lagangian Euaion

More information

( ) exp i ω b ( ) [ III-1 ] exp( i ω ab. exp( i ω ba

( ) exp i ω b ( ) [ III-1 ] exp( i ω ab. exp( i ω ba THE INTEACTION OF ADIATION AND MATTE: SEMICLASSICAL THEOY PAGE 26 III. EVIEW OF BASIC QUANTUM MECHANICS : TWO -LEVEL QUANTUM SYSTEMS : The lieaue of quanum opics and lase specoscop abounds wih discussions

More information

WORK POWER AND ENERGY Consevaive foce a) A foce is said o be consevaive if he wok done by i is independen of pah followed by he body b) Wok done by a consevaive foce fo a closed pah is zeo c) Wok done

More information

In the complete model, these slopes are ANALYSIS OF VARIANCE FOR THE COMPLETE TWO-WAY MODEL. (! i+1 -! i ) + [(!") i+1,q - [(!

In the complete model, these slopes are ANALYSIS OF VARIANCE FOR THE COMPLETE TWO-WAY MODEL. (! i+1 -! i ) + [(!) i+1,q - [(! ANALYSIS OF VARIANCE FOR THE COMPLETE TWO-WAY MODEL The frs hng o es n wo-way ANOVA: Is here neracon? "No neracon" means: The man effecs model would f. Ths n urn means: In he neracon plo (wh A on he horzonal

More information

T h e C S E T I P r o j e c t

T h e C S E T I P r o j e c t T h e P r o j e c t T H E P R O J E C T T A B L E O F C O N T E N T S A r t i c l e P a g e C o m p r e h e n s i v e A s s es s m e n t o f t h e U F O / E T I P h e n o m e n o n M a y 1 9 9 1 1 E T

More information

Machine Learning. Spectral Clustering. Lecture 23, April 14, Reading: Eric Xing 1

Machine Learning. Spectral Clustering. Lecture 23, April 14, Reading: Eric Xing 1 Machne Leanng -7/5 7/5-78, 78, Spng 8 Spectal Clusteng Ec Xng Lectue 3, pl 4, 8 Readng: Ec Xng Data Clusteng wo dffeent ctea Compactness, e.g., k-means, mxtue models Connectvty, e.g., spectal clusteng

More information

Handling Fuzzy Constraints in Flow Shop Problem

Handling Fuzzy Constraints in Flow Shop Problem Handlng Fuzzy Consans n Flow Shop Poblem Xueyan Song and Sanja Peovc School of Compue Scence & IT, Unvesy of Nongham, UK E-mal: {s sp}@cs.no.ac.uk Absac In hs pape, we pesen an appoach o deal wh fuzzy

More information

PHYS 705: Classical Mechanics. Derivation of Lagrange Equations from D Alembert s Principle

PHYS 705: Classical Mechanics. Derivation of Lagrange Equations from D Alembert s Principle 1 PHYS 705: Classcal Mechancs Devaton of Lagange Equatons fom D Alembet s Pncple 2 D Alembet s Pncple Followng a smla agument fo the vtual dsplacement to be consstent wth constants,.e, (no vtual wok fo

More information

Servomechanism Design

Servomechanism Design Sevomechanism Design Sevomechanism (sevo-sysem) is a conol sysem in which he efeence () (age, Se poin) changes as ime passes. Design mehods PID Conol u () Ke P () + K I ed () + KDe () Sae Feedback u()

More information

Chapter 6 DETECTION AND ESTIMATION: Model of digital communication system. Fundamental issues in digital communications are

Chapter 6 DETECTION AND ESTIMATION: Model of digital communication system. Fundamental issues in digital communications are Chaper 6 DCIO AD IMAIO: Fndaenal sses n dgal concaons are. Deecon and. saon Deecon heory: I deals wh he desgn and evalaon of decson ang processor ha observes he receved sgnal and gesses whch parclar sybol

More information

β A Constant-G m Biasing

β A Constant-G m Biasing p 2002 EE 532 Anal IC Des II Pae 73 Cnsan-G Bas ecall ha us a PTAT cuen efeence (see p f p. 66 he nes) bas a bpla anss pes cnsan anscnucance e epeaue (an als epenen f supply lae an pcess). Hw h we achee

More information

Sections 3.1 and 3.4 Exponential Functions (Growth and Decay)

Sections 3.1 and 3.4 Exponential Functions (Growth and Decay) Secions 3.1 and 3.4 Eponenial Funcions (Gowh and Decay) Chape 3. Secions 1 and 4 Page 1 of 5 Wha Would You Rahe Have... $1million, o double you money evey day fo 31 days saing wih 1cen? Day Cens Day Cens

More information

Lagrangian & Hamiltonian Mechanics:

Lagrangian & Hamiltonian Mechanics: XII AGRANGIAN & HAMITONIAN DYNAMICS Iouco Hamlo aaoal Pcple Geealze Cooaes Geealze Foces agaga s Euao Geealze Momea Foces of Cosa, agage Mulples Hamloa Fucos, Cosevao aws Hamloa Dyamcs: Hamlo s Euaos agaga

More information

Notes on the stability of dynamic systems and the use of Eigen Values.

Notes on the stability of dynamic systems and the use of Eigen Values. Noes on he sabl of dnamc ssems and he use of Egen Values. Source: Macro II course noes, Dr. Davd Bessler s Tme Seres course noes, zarads (999) Ineremporal Macroeconomcs chaper 4 & Techncal ppend, and Hamlon

More information

3.2 Models for technical systems

3.2 Models for technical systems onrol Laboraory 3. Mahemacal Moelng 3. Moels for echncal sysems 3.. Elecrcal sysems Fg. 3. shows hree basc componens of elecrcal crcs. Varables = me, = volage [V], = crren [A] omponen parameers R = ressance

More information

CHAPTER 13 LAGRANGIAN MECHANICS

CHAPTER 13 LAGRANGIAN MECHANICS CHAPTER 3 AGRANGIAN MECHANICS 3 Inoucon The usual way of usng newonan mechancs o solve a poblem n ynamcs s fs of all o aw a lage, clea agam of he sysem, usng a ule an a compass Then mak n he foces on he

More information

Energy in Closed Systems

Energy in Closed Systems Enegy n Closed Systems Anamta Palt palt.anamta@gmal.com Abstact The wtng ndcates a beakdown of the classcal laws. We consde consevaton of enegy wth a many body system n elaton to the nvese squae law and

More information

P a g e 3 6 of R e p o r t P B 4 / 0 9

P a g e 3 6 of R e p o r t P B 4 / 0 9 P a g e 3 6 of R e p o r t P B 4 / 0 9 p r o t e c t h um a n h e a l t h a n d p r o p e r t y fr om t h e d a n g e rs i n h e r e n t i n m i n i n g o p e r a t i o n s s u c h a s a q u a r r y. J

More information

A Compact Representation of Spatial Correlation in MIMO Radio Channels

A Compact Representation of Spatial Correlation in MIMO Radio Channels A Compac epesenaon of Spaal Coelaon n MIMO ado Channels A. van Zels Endhoven Unves of echnolog P.O. Box 53 5600 MB Endhoven he Nehelands e-mal: A.v.Zels@ue.nl and Agee Ssems P.O. Box 755 3430 A Neuwegen

More information

How to Obtain Desirable Transfer Functions in MIMO Systems Under Internal Stability Using Open and Closed Loop Control

How to Obtain Desirable Transfer Functions in MIMO Systems Under Internal Stability Using Open and Closed Loop Control How to Obtain Desiable ansfe Functions in MIMO Sstems Unde Intenal Stabilit Using Open and losed Loop ontol echnical Repot of the ISIS Goup at the Univesit of Note Dame ISIS-03-006 June, 03 Panos J. Antsaklis

More information

p E p E d ( ) , we have: [ ] [ ] [ ] Using the law of iterated expectations, we have:

p E p E d ( ) , we have: [ ] [ ] [ ] Using the law of iterated expectations, we have: Poblem Se #3 Soluons Couse 4.454 Maco IV TA: Todd Gomley, gomley@m.edu sbued: Novembe 23, 2004 Ths poblem se does no need o be uned n Queson #: Sock Pces, vdends and Bubbles Assume you ae n an economy

More information

2/20/2013. EE 101 Midterm 2 Review

2/20/2013. EE 101 Midterm 2 Review //3 EE Mderm eew //3 Volage-mplfer Model The npu ressance s he equalen ressance see when lookng no he npu ermnals of he amplfer. o s he oupu ressance. I causes he oupu olage o decrease as he load ressance

More information

I. G. PETROVSKY LECTURES ON PARTIAL DIFFERENTIAL EQUATIONS FIRST ENGLISH EDITION 1954

I. G. PETROVSKY LECTURES ON PARTIAL DIFFERENTIAL EQUATIONS FIRST ENGLISH EDITION 1954 I. G. PETROVSKY LECTURES ON PARTIAL DIFFERENTIAL EQUATIONS FIRST ENGLISH EDITION 954 CONTENTS Forewor b R. Coran Translaor's Noe b Abe Shenzer Preface Chaper I. Inrocon. Classfcaon of eqaons. Defnons.

More information

A Trajectory Planner for Autonomous Structural Assembly

A Trajectory Planner for Autonomous Structural Assembly A Tajeco Planne fo Aonomos cal Assembl ahane L. R. cghan * Unves of alan, ollege Pak, alan, 7 Aonomos oboc sace conscon eqes obs schelng ones o ecomose an oe assembl acves an an effcen ajeco lanne o s

More information

John Geweke a and Gianni Amisano b a Departments of Economics and Statistics, University of Iowa, USA b European Central Bank, Frankfurt, Germany

John Geweke a and Gianni Amisano b a Departments of Economics and Statistics, University of Iowa, USA b European Central Bank, Frankfurt, Germany Herarchcal Markov Normal Mxure models wh Applcaons o Fnancal Asse Reurns Appendx: Proofs of Theorems and Condonal Poseror Dsrbuons John Geweke a and Gann Amsano b a Deparmens of Economcs and Sascs, Unversy

More information

Fuzzy Control of Inverted Robot Arm with Perturbed Time-Delay Affine Takagi-Sugeno Fuzzy Model

Fuzzy Control of Inverted Robot Arm with Perturbed Time-Delay Affine Takagi-Sugeno Fuzzy Model 7 IEEE Inenaonal Confeence on Robocs an Auomaon Roma Ialy -4 Al 7 FD5. Fuzzy Conol of Invee Robo Am wh Peube me-delay Affne akag-sugeno Fuzzy Moel Wen-Je Chang We-Han Huang an We Chang Absac A sably analyss

More information

THEORETICAL AUTOCORRELATIONS. ) if often denoted by γ. Note that

THEORETICAL AUTOCORRELATIONS. ) if often denoted by γ. Note that THEORETICAL AUTOCORRELATIONS Cov( y, y ) E( y E( y))( y E( y)) ρ = = Var( y) E( y E( y)) =,, L ρ = and Cov( y, y ) s ofen denoed by whle Var( y ) f ofen denoed by γ. Noe ha γ = γ and ρ = ρ and because

More information

Chapter 13 - Universal Gravitation

Chapter 13 - Universal Gravitation Chapte 3 - Unesal Gataton In Chapte 5 we studed Newton s thee laws of moton. In addton to these laws, Newton fomulated the law of unesal gataton. Ths law states that two masses ae attacted by a foce gen

More information

Control Volume Derivation

Control Volume Derivation School of eospace Engineeing Conol Volume -1 Copyigh 1 by Jey M. Seizman. ll ighs esee. Conol Volume Deiaion How o cone ou elaionships fo a close sysem (conol mass) o an open sysem (conol olume) Fo mass

More information