Adaptive Control 2

download Adaptive Control 2

of 30

Transcript of Adaptive Control 2

  • 8/9/2019 Adaptive Control 2

    1/30

    ADAPTIVE CONTROL

    SYSTEMS

  • 8/9/2019 Adaptive Control 2

    2/30

    MRAC MODEL REFERENCE ADAPTIVE CONTROL

    SYSTEM It can be considered as an ADAPTIVE SERVO

    SYSTEM. It consists of two oo!s inner oo! "o#ter oo! It consists of a reference $ode in o#ter oo!. Ordinar% feed bac& is caed inner oo!.

    Para$eters are ad'#sted on basis of feedbac&fro$ t(e error. Error is t(e difference between !rod#ced o#t!#t

    " reference $ode )a#e.

  • 8/9/2019 Adaptive Control 2

    3/30

    Adjustment of system parameters

    in a MRAC can be obtained in twoways.

    GRADIENT METHD !MIT R"#E$

    #%&N' (TA)I#IT% THER%

  • 8/9/2019 Adaptive Control 2

    4/30

    MODEL

    CONTROLLER PLANT

    ADJUSTMENT

    MECHANISM

    *C

    YM

    * Y

    CONTROLLER PARAMETERS

    )#C* DIAGRAM + MRAC

  • 8/9/2019 Adaptive Control 2

    5/30

    MRAC IS COMPOSED OF

    Pant containin+ #n&nown !ara$eters

    Reference $ode Ad'#stabe !ara$eters containin+ contro

    aw or oss aw

    Ordinar% feed bac& oo!

  • 8/9/2019 Adaptive Control 2

    6/30

    MIT R*LE

    It is de)eo!ed b% INSTR*MANTATION LA,ORATORY

    at MIT.

    To consider MIT r#e we #se a cosed oo! res!onse inw(ic( controer (as one ad'#stabe !ara$eter -.

    T(e desired cosed oo! res!onse is s!ecified b% a

    $ode w(ose o#t!#t is %$.

    T(en error e %$/ %. 0% is ori+ina o#t!#t1.

  • 8/9/2019 Adaptive Control 2

    7/30

    Error criteria seected (ere is

    ,!-$/01 !e$1

    And t(is oss f#nction is to be $ini$i2ed to $a&e t(e s%ste$ controed.To ac(ie)e t(is t(e !ara$eters are c(an+ed in t(e direction of ne+ati)e

    +radient of 3.

    d- 23 4, 2 3e 4e

    dt d- 4-

    T(is is caedGRADIENT orMIT ru5e.

    3 adaptation 6ain

    4e0 4- sensiti7ity deri7ati7e!informs 8ow error isinf5uenced by -

  • 8/9/2019 Adaptive Control 2

    8/30

    T(e oss f#nction can aso be c(osen as

    ,!-$9e9

    d- 2 3 4e si6n e :GRADIENT METHD;

    dt 4-

    first MRAC is imp5emented by t8is function

    d- 2 3 si6n 4e si6n !e$

    dt 4- : si6n2si6n a56orit8m;

    used in te5ecommunication w8ere simp5e

    imp5ementation < fast computin6 are re=uired

  • 8/9/2019 Adaptive Control 2

    9/30

    +or mu5ti7ariab5e systems

    - is considered as 7ector

    4e0 4- is considered as 6radient of t8e

    error wit8 respect to t8e parameter

  • 8/9/2019 Adaptive Control 2

    10/30

    A&ICATIN( +R MIT

    R"#EADAPTATION OF A FEED FOR4ARD 5AIN

    MRAS FOR A FIRST ORDER SYSTEM

  • 8/9/2019 Adaptive Control 2

    11/30

    ADAPTATION OF A FEED FOR4ARD

    5AIN

    PRO,LEM 6Ad'#st$ent of feed forward +ain

    ASS*MPTIONS6 Process is inear wit( t(e transfer f#nction 750s1 50s1 is &nown 7 is #n&nown !ara$eterDESIRED CONDITION6Transfer f#nction 5$0s1 s(o#d be e8#a to 7o50s1

    7ois +i)en constant

  • 8/9/2019 Adaptive Control 2

    12/30

    9 :

    9 750S1

    ;

    ?

    - " "C

    %m

    MRAC +R AD,"(TMENT + +EED+R>ARD GAIN )% MIT R"#E

  • 8/9/2019 Adaptive Control 2

    13/30

    T(e constant kfor t(is !ant is #n&nown. @owe)er a

    reference $ode can be for$ed wit( a desired )a#e of k

    and t(ro#+( ada!tation of a feedforward +ain t(e

    res!onse of t(e !ant can be $ade to $atc( t(is $ode.

    T(e reference $ode is t(erefore c(osen as t(e !ant

    $#ti!ied b% a desired constant ko.

    T(e cost f#nction c(osen (ere is

  • 8/9/2019 Adaptive Control 2

    14/30

    T(e error is t(en restated in ter$s of t(e transfer f#nctions

    $#ti!ied b% t(eir in!#ts.

    As can be seen t(is eB!ression for t(e error

    contains t(e !ara$eter thetaw(ic( is to be #!dated.

    To deter$ine t(e #!date r#e t(e sensiti)it%

    deri)ati)e is cac#ated and restated in ter$s of t(e

    $ode o#!#t6

  • 8/9/2019 Adaptive Control 2

    15/30

    Fina% t(e MIT r#e is a!!ied to +i)e an eB!ression for#!datin+ theta. T(e constants kand koare co$bined into

    gamma.

    To t#ne t(is s%ste$ t(e )a#es of koand gammacan be

    )aried.

  • 8/9/2019 Adaptive Control 2

    16/30

    NTE( N DE(IGN >ITH MIT

    It is i$!ortant to note t(at t(e MIT r#e b% itsef does not

    +#arantee con)er+ence or stabiit%.An MRAC desi+ned #sin+ t(e MIT r#e is )er% sensiti)e

    to t(e a$!it#des of t(e si+nas.

    As a +enera r#e t(e )a#e of gammais &e!t s$a.T#nin+ of gammais cr#cia to t(e ada!tation rate and

    stabiit% of t(e controer.

  • 8/9/2019 Adaptive Control 2

    17/30

    IF 50S1=S?

    *C is sin#soida si+na wit( fre8#enc% rad=s

    7

    7

    Var% )a#es of ARD GAIN CNTR# "(ING #%A&"N' R"#E

  • 8/9/2019 Adaptive Control 2

    26/30

  • 8/9/2019 Adaptive Control 2

    27/30

    T8e error e=uations are

  • 8/9/2019 Adaptive Control 2

    28/30

  • 8/9/2019 Adaptive Control 2

    29/30

    T8e 7a5ue of d70dt is :w8ic8 is 7e semidefinite;

    As time deri7ati7e of #ypano7 function is ne6ati7e

    semi definite.

    )y usin6 5emma of 5ypano7 we can s8ow error

    6oes to Fero.

    (tt!6==www contro (#t fi=7#rssit=AS;GH =#ennot=#e! !df

    http://www.control.hut.fi/Kurssit/AS-74.185/luennot/lu5ep.pdfhttp://www.control.hut.fi/Kurssit/AS-74.185/luennot/lu5ep.pdf
  • 8/9/2019 Adaptive Control 2

    30/30

    (tt!6==www.contro.(#t.fi=7#rssit=AS GH.=#ennot=#e!.!df

    (tt!6==www.contro.(#t.fi=7#rssit=AS;GH.=#ennot=#Je!.!df

    (tt!6==www.contro.t(.se=KFRT=EBercises=eBHso.!df

    (tt!6==www.contro.t(.se=GEFRT=EBercises=eBso.!df

    (tt!6==www.i+i.t#+ra2.at=(e$#t=Presentations=Ada!ti)eContro

    .(t$

    (tt!6==www.irs.ctr.titec(.ac.'!=Kdra=icH=ecH.!df

    (tt!6==www.irs.ctr.titec(.ac.'!=Kdra=icH=ecHJ.!df

    (tt!6==$c(ab.ee.n#s.ed#.s+=EB!eri$ent=Man#as=EEH=ada

    ! !df

    http://www.control.hut.fi/Kurssit/AS-74.185/luennot/lu5ep.pdfhttp://www.control.hut.fi/Kurssit/AS-74.185/luennot/lu6ep.pdfhttp://www.control.lth.se/~FRT050/Exercises/ex4sol.pdfhttp://www.control.lth.se/~FRT050/Exercises/ex5sol.pdfhttp://www.igi.tugraz.at/helmut/Presentations/AdaptiveControl.htmlhttp://www.igi.tugraz.at/helmut/Presentations/AdaptiveControl.htmlhttp://www.irs.ctrl.titech.ac.jp/~dkura/ic2004/lec405.pdfhttp://www.irs.ctrl.titech.ac.jp/~dkura/ic2004/lec406.pdfhttp://mchlab.ee.nus.edu.sg/Experiment/Manuals/EE5140/adap1.pdfhttp://mchlab.ee.nus.edu.sg/Experiment/Manuals/EE5140/adap1.pdfhttp://mchlab.ee.nus.edu.sg/Experiment/Manuals/EE5140/adap1.pdfhttp://mchlab.ee.nus.edu.sg/Experiment/Manuals/EE5140/adap1.pdfhttp://www.irs.ctrl.titech.ac.jp/~dkura/ic2004/lec406.pdfhttp://www.irs.ctrl.titech.ac.jp/~dkura/ic2004/lec405.pdfhttp://www.igi.tugraz.at/helmut/Presentations/AdaptiveControl.htmlhttp://www.igi.tugraz.at/helmut/Presentations/AdaptiveControl.htmlhttp://www.control.lth.se/~FRT050/Exercises/ex5sol.pdfhttp://www.control.lth.se/~FRT050/Exercises/ex4sol.pdfhttp://www.control.hut.fi/Kurssit/AS-74.185/luennot/lu6ep.pdfhttp://www.control.hut.fi/Kurssit/AS-74.185/luennot/lu5ep.pdf