Adaptive Dynamic Programming for Control: Algorithms and Stability

431 Pages·2013·6.079 MB·English

by  Huaguang Zhang, Derong Liu, Yanhong Luo, Ding Wang (auth.)

Communications and Control Engineering
Forfurthervolumes:
www.springer.com/series/61
Huaguang Zhang (cid:2) Derong Liu (cid:2) Yanhong Luo (cid:2)
Ding Wang
Adaptive Dynamic
Programming
for Control
Algorithms and Stability
HuaguangZhang YanhongLuo
CollegeofInformationScienceEngin. CollegeofInformationScienceEngin.
NortheasternUniversity NortheasternUniversity
Shenyang Shenyang
People’sRepublicofChina People’sRepublicofChina
DerongLiu DingWang
InstituteofAutomation,Laboratory InstituteofAutomation,Laboratory
ofComplexSystems ofComplexSystems
ChineseAcademyofSciences ChineseAcademyofSciences
Beijing Beijing
People’sRepublicofChina People’sRepublicofChina
ISSN0178-5354 CommunicationsandControlEngineering
ISBN978-1-4471-4756-5 ISBN978-1-4471-4757-2(eBook)
DOI10.1007/978-1-4471-4757-2
SpringerLondonHeidelbergNewYorkDordrecht
LibraryofCongressControlNumber:2012955288
©Springer-VerlagLondon2013
Thisworkissubjecttocopyright.AllrightsarereservedbythePublisher,whetherthewholeorpartof
thematerialisconcerned,specificallytherightsoftranslation,reprinting,reuseofillustrations,recitation,
broadcasting,reproductiononmicrofilmsorinanyotherphysicalway,andtransmissionorinformation
storageandretrieval,electronicadaptation,computersoftware,orbysimilarordissimilarmethodology
nowknownorhereafterdeveloped.Exemptedfromthislegalreservationarebriefexcerptsinconnection
with reviews or scholarly analysis or material supplied specifically for the purpose of being entered
and executed on a computer system, for exclusive use by the purchaser of the work. Duplication of
this publication or parts thereof is permitted only under the provisions of the Copyright Law of the
Publisher’slocation,initscurrentversion,andpermissionforusemustalwaysbeobtainedfromSpringer.
PermissionsforusemaybeobtainedthroughRightsLinkattheCopyrightClearanceCenter.Violations
areliabletoprosecutionundertherespectiveCopyrightLaw.
Theuseofgeneraldescriptivenames,registerednames,trademarks,servicemarks,etc.inthispublication
doesnotimply,evenintheabsenceofaspecificstatement,thatsuchnamesareexemptfromtherelevant
protectivelawsandregulationsandthereforefreeforgeneraluse.
Whiletheadviceandinformationinthisbookarebelievedtobetrueandaccurateatthedateofpub-
lication,neithertheauthorsnortheeditorsnorthepublishercanacceptanylegalresponsibilityforany
errorsoromissionsthatmaybemade.Thepublishermakesnowarranty,expressorimplied,withrespect
tothematerialcontainedherein.
Printedonacid-freepaper
SpringerispartofSpringerScience+BusinessMedia(www.springer.com)
Preface
Background ofThisBook
Optimal control, once thought of as one of the principal and complex domains in
the control field, has been studied extensively in both science and engineering for
severaldecades.Asisknown,dynamicalsystemsareubiquitousinnatureandthere
exist many methods to design stable controllers for dynamical systems. However,
stability is only a bare minimum requirement in the design of a system. Ensuring
optimalityguaranteesthestabilityofnonlinearsystems.Asanextensionofthecal-
culus of variations, optimal control theory is a mathematical optimization method
forderivingcontrolpolicies.Dynamicprogrammingisaveryusefultoolinsolving
optimizationandoptimalcontrolproblemsbyemployingtheprincipleofoptimality.
However, it is often computationally untenable to run true dynamic programming
duetothewell-known“curseofdimensionality”.Hence,theadaptivedynamicpro-
gramming (ADP) method was first proposed by Werbos in 1977. By building a
system,called“critic”,toapproximatethecostfunctionindynamicprogramming,
onecanobtaintheapproximateoptimalcontrolsolutiontodynamicprogramming.
In recent years, ADP algorithms have gained much attention from researchers in
controlfields.However,withthedevelopmentofADPalgorithms,moreandmore
peoplewanttoknowtheanswerstothefollowingquestions:
(1) AreADPalgorithmsconvergent?
(2) Canthealgorithmstabilizeanonlinearplant?
(3) Canthealgorithmberunon-line?
(4) Canthealgorithmbeimplementedinafinitetimehorizon?
(5) Iftheanswertothefirstquestionispositive,thesubsequentquestionsarewhere
thealgorithmconvergesto,andhowlargetheerroris.
Before ADP algorithms can be applied to real plants, these questions need to
beansweredfirst.Throughoutthisbook,wewillstudyallthesequestionsandgive
specificanswerstoeachquestion.
v
vi Preface
WhyThisBook?
AlthoughlotsofmonographsonADPhaveappeared,thepresentbookhasunique
features,whichdistinguishitfromothers.
First,thetypesofsysteminvolvedinthismonographareratherextensive.From
thepointofviewofmodels,onecanfindaffinenonlinearsystems,non-affinenon-
linearsystems,switchednonlinearsystems,singularlyperturbedsystemsandtime-
delaynonlinearsystemsinthisbook;thesearethemainmathematicalmodelsinthe
controlfields.
Second,sincethemonographisasummaryofrecentresearchworksoftheau-
thors, the methods presented here for stabilizing, tracking, and games, which to a
greatdegreebenefitfromoptimalcontroltheory,aremoreadvancedthanthoseap-
pearinginintroductorybooks.Forexample,thedualheuristicprogrammingmethod
isusedtostabilizeaconstrainednonlinearsystem,withconvergenceproof;adata-
based robust approximate optimal controller is designed based on simultaneous
weightupdatingoftwonetworks;andasinglenetworkschemeisproposedtosolve
thenon-zero-sumgameforaclassofcontinuous-timesystems.
Last but not least, some rather unique contributions are included in this mono-
graph. One notable feature is the implementation of finite horizon optimal control
fordiscrete-timenonlinearsystems,whichcanobtainsuboptimalcontrolsolutions
within a fixed finite number of control steps. Most existing results in other books
discussonlytheinfinitehorizoncontrol,whichisnotpreferredinreal-worldappli-
cations.Besidesthisfeature,anothernotablefeatureisthatapairofmixedoptimal
policies is developed to solve nonlinear games for the first time when the saddle
pointdoesnotexist.Meanwhile,forthesituationthatthesaddlepointexists,exis-
tenceconditionsofthesaddlepointareavoided.
TheContentof ThisBook
Thebookinvolvestenchapters.Asimpliedbythebooktitle,themaincontentofthe
bookiscomposedofthreeparts;thatis,optimalfeedbackcontrol,nonlineargames,
andrelatedapplicationsofADP.Inthepartonoptimalfeedbackcontrol,theedge-
cutting results on ADP-based infinite horizon and finite horizon feedback control,
including stabilization control, and tracking control are presented in a systematic
manner. In the part on nonlinear games, both zero-sum game and non-zero-sum
games are studied. For the zero-sum game, it is proved for the first time that the
iterative policies converge to the mixed optimal solutions when the saddle point
doesnotexist.Forthenon-zero-sumgame,asinglenetworkisproposedtoseekthe
Nash equilibrium for the first time. In the part of applications, a self-learning call
admissioncontrolschemeisproposedforCDMAcellularnetworks,andmeanwhile
an engine torque and air-fuel ratio control scheme is studied in detail, based on
ADP.
In Chap. 1, a brief introduction to the background and development of ADP
is provided. The review begins with the origin of ADP, and the basic structures
Preface vii
andalgorithmdevelopmentarenarratedinchronologicalorder.Afterthat,weturn
attentiontocontrolproblemsbasedonADP.Wepresentthissubjectregardingtwo
aspects: feedback control based on ADP and nonlinear games based on ADP. We
mention a few iterative algorithms from recent literature and point out some open
problemsineachcase.
InChap.2,theoptimalstatefeedbackcontrolproblemisstudiedbasedonADP
for both infinite horizon and finite horizon. Three different structures of ADP are
utilized to solve the optimal state feedback control strategies, respectively. First,
consideringaclassofaffineconstrainedsystems,anewDHPmethodisdeveloped
tostabilizethesystem,withconvergenceproof.Then,duetothespecialadvantages
ofGDHPstructure,anewoptimalcontrolschemeisdevelopedwithdiscountedcost
functional. Moreover, based on a least-square successive approximation method,
a series of GHJB equations are solved to obtain the optimal control solutions. Fi-
nally,anovelfinite-horizonoptimalcontrolschemeisdevelopedtoobtainthesub-
optimal control solutions within a fixed finite number of control steps. Compared
withtheexistingresultsintheinfinite-horizoncase,thepresentfinite-horizonopti-
malcontrollerispreferredinreal-worldapplications.
Chapter 3 presents some direct methods for solving the closed-loop optimal
tracking control problem for discrete-time systems. Considering the fact that the
performanceindexfunctionsofoptimaltrackingcontrolproblemsarequitediffer-
ent from those of optimal state feedback control problems, a new type of perfor-
mance index function is defined. The methods are mainly based on iterative HDP
andGDHPalgorithms.Wefirststudytheoptimaltrackingcontrolproblemofaffine
nonlinearsystems,andafterthatwestudytheoptimaltrackingcontrolproblemof
non-affinenonlinearsystems.Itisnoticedthatmostreal-worldsystemsneedtobe
effectively controlled within a finite time horizon. Hence, based on the above re-
sults, we further study the finite-horizon optimal tracking control problem, using
theADPapproachinthelastpartofChap.3.
In Chap. 4, the optimal state feedback control problems of nonlinear systems
withtimedelaysarestudied.Ingeneral,theoptimalcontrolfortime-delaysystems
isaninfinite-dimensionalcontrolproblem,whichisverydifficulttosolve;thereare
presentlynogoodmethodsfordealingwiththisproblem.Inthischapter,theopti-
malstatefeedbackcontrolproblemsofnonlinearsystemswithtimedelaysbothin
statesandcontrolsareinvestigated.Byintroducingadelaymatrixfunction,theex-
plicitexpressionoftheoptimalcontrolfunctioncanbeobtained.Next,fornonlinear
time-delay systems with saturating actuators, we further study the optimal control
problem using a non-quadratic functional, where two optimization processes are
developed for searching the optimal solutions. The above two results are for the
infinite-horizon optimal control problem. To the best of our knowledge, there are
no results on the finite-horizon optimal control of nonlinear time-delay systems.
Hence, in the last part of this chapter, a novel optimal control strategy is devel-
oped to solve the finite-horizon optimal control problem for a class of time-delay
systems.
InChap.5,theoptimaltrackingcontrolproblemsofnonlinearsystemswithtime
delays are studied using the HDP algorithm. First, the HJB equation for discrete
viii Preface
time-delaysystemsisderivedbasedonstateerrorandcontrolerror.Then,anovel
iterativeHDPalgorithmcontainingtheiterationsofstate,controllaw,andcostfunc-
tional is developed. We also give the convergence proof for the present iterative
HDP algorithm. Finally, two neural networks, i.e., the critic neural network and
theactionneuralnetwork,areusedtoapproximatethevaluefunctionandthecor-
responding control law, respectively. It is the first time that the optimal tracking
control problem of nonlinear systems with time delays is solved using the HDP
algorithm.
In Chap. 6, we focus on the design of controllers for continuous-time systems
via the ADP approach. Although many ADP methods have been proposed for
continuous-timesystems,asuitableframeworkinwhichtheoptimalcontrollercan
be designed for a class of general unknown continuous-time systems still has not
beendeveloped.Inthefirstpartofthischapter,wedevelopanewschemetodesign
optimalrobusttrackingcontrollersforunknowngeneralcontinuous-timenonlinear
systems.Themeritofthepresentmethodisthatwerequireonlytheavailabilityof
input/outputdata,insteadofanexactsystemmodel.Theobtainedcontrolinputcan
beguaranteedtobeclosetotheoptimalcontrolinputwithinasmallbound.Inthe
second part of the chapter, a novel ADP-based robust neural network controller is
developedforaclassofcontinuous-timenon-affinenonlinearsystems,whichisthe
first attempt to extend the ADP approach to continuous-time non-affine nonlinear
systems.
In Chap. 7, several special optimal feedback control schemes are investigated.
Inthefirstpart,theoptimalfeedbackcontrolproblemofaffinenonlinearswitched
systems is studied. To seek optimal solutions, a novel two-stage ADP method is
developed. The algorithm can be divided into two stages: first, for each possible
mode,calculatetheassociatedvaluefunction,andthenselecttheoptimalmodefor
eachstate.Inthesecondandthirdparts,thenear-optimalcontrollersfornonlinear
descriptor systems and singularly perturbed systems are solved by iterative DHP
andHDPalgorithms,respectively.Inthefourthpart,thenear-optimalstate-feedback
controlproblemofnonlinearconstraineddiscrete-timesystemsissolvedviaasingle
network ADP algorithm. At each step of the iterative algorithm, a neural network
isutilizedtoapproximatethecostatefunction,andthentheoptimalcontrolpolicy
of the system can be computed directly according to the costate function, which
removestheactionnetworkappearingintheordinaryADPstructure.
Gametheoryisconcernedwiththestudyofdecisionmakinginasituationwhere
two or more rational opponents are involved under conditions of conflicting inter-
ests.InChap.8,zero-sumgamesareinvestigatedfordiscrete-timesystemsbasedon
themodel-freeADPmethod.First,aneffectivedata-basedoptimalcontrolschemeis
developedviatheiterativeADPalgorithmtofindtheoptimalcontrollerofaclassof
discrete-timezero-sumgamesforRoessertype2-Dsystems.Sincetheexactmod-
els of many 2-D systems cannot be obtained inherently, the iterative ADP method
isexpectedtoavoidtherequirementofexactsystemmodels.Second,adata-based
optimaloutputfeedbackcontrollerisdevelopedforsolvingthezero-sumgamesof
aclassofdiscrete-timesystems,whosemeritisthatknowledgeofthemodelofthe
systemisnotrequired,northeinformationofsystemstates.
Preface ix
In Chap. 9, nonlinear game problems are investigated for continuous-time sys-
tems,includinginfinitehorizonzero-sumgames,finitehorizonzero-sumgamesand
non-zero-sumgames.First,forthesituationsthatthesaddlepointexists,theADP
techniqueisusedtoobtaintheoptimalcontrolpairiteratively.Thepresentapproach
makestheperformanceindexfunctionreachthesaddlepointofthezero-sumdiffer-
ential games, while complex existence conditions of the saddle point are avoided.
Forthesituationsthatthesaddlepointdoesnotexist,themixedoptimalcontrolpair
isobtainedtomaketheperformanceindexfunctionreachthemixedoptimum.Then,
finitehorizonzero-sumgamesfor aclassofnonaffinenonlinearsystemsarestud-
ied.Moreover,besidesthezero-sumgames,thenon-zero-sumdifferentialgamesare
studiedbasedonsingle networkADPalgorithm.For zero-sumdifferentialgames,
twoplayersworkonacostfunctionaltogetherandminimaxit.However,fornon-
zero-sumgames,thecontrolobjectiveistofindasetofpoliciesthatguaranteethe
stabilityofthesystemandminimizetheindividualperformancefunctiontoyielda
Nashequilibrium.
InChap.10,theoptimalcontrolproblemsofmodernwirelessnetworksandauto-
motiveenginesarestudiedbyusingADPmethods.Inthefirstpart,anovellearning
controlarchitectureis proposedbasedonadaptivecriticdesigns/ADP,withonlya
singlemoduleinsteadoftwoorthreemodules.Thechoiceofutilityfunctionforthe
presentself-learningcontrolschememakesthepresentlearningprocessmuchmore
efficient than existing learning control methods. The call admission controller can
performlearninginrealtimeaswellasinoff-lineenvironments,andthecontroller
improvesitsperformanceasitgainsmoreexperience.Inthesecondpart,anADP-
basedlearningalgorithmisdesignedaccordingtocertaincriteriaandcalibratedfor
vehicleoperationover theentire operatingregime.The algorithmis optimizedfor
the engine in terms of performance, fuel economy, and tailpipe emissions through
asignificanteffortinresearchanddevelopmentandcalibrationprocesses.Afterthe
controller has learned to provide optimal control signals under various operating
conditionsoff-lineoron-line,itisappliedtoperformthetaskofenginecontrolin
real time. The performance of the controller can be further refined and improved
throughcontinuouslearninginreal-timevehicleoperations.
Acknowledgments
Theauthorswouldliketoacknowledgethehelpandencouragementtheyreceived
during the course of writing this book. A great deal of the materials presented in
this book is based on the research that we conducted with several colleagues and
formerstudents,includingQ.L.Wei,Y.Zhang,T.Huang,O.Kovalenko,L.L.Cui,
X.Zhang,R.Z.SongandN.Cao.WewishtoacknowledgeespeciallyDr.J.L.Zhang
andDr.C.B.Qinfortheirhardworkonthisbook.Theauthorsalsowishtothank
Prof.R.E.Bellman,Prof.D.P.Bertsekas,Prof.F.L.Lewis,Prof.J.SiandProf.S.Ja-
gannathan for their excellent books on the theory of optimal control and adaptive
x Preface
dynamicprogramming.WeareverygratefultotheNationalNaturalScienceFoun-
dation of China (50977008, 60904037, 61034005, 61034002, 61104010), the Sci-
enceandTechnologyResearchProgramofTheEducationDepartmentofLiaoning
Province(LT2010040),whichprovidednecessaryfinancialsupportforwritingthis
book.
Shenyang,China HuaguangZhang
Beijing,China DerongLiu
Chicago,USA YanhongLuo
DingWang
Contents
1 Overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.1 ChallengesofDynamicProgramming . . . . . . . . . . . . . . . . 1
1.2 BackgroundandDevelopmentofAdaptiveDynamicProgramming 3
1.2.1 BasicStructuresofADP . . . . . . . . . . . . . . . . . . . 4
1.2.2 RecentDevelopmentsofADP . . . . . . . . . . . . . . . . 6
1.3 FeedbackControlBasedonAdaptiveDynamicProgramming . . . 11
1.4 Non-linearGamesBasedonAdaptiveDynamicProgramming . . . 17
1.5 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
2 OptimalStateFeedbackControlforDiscrete-TimeSystems . . . . . 27
2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
2.2 Infinite-HorizonOptimalStateFeedbackControlBasedonDHP . 27
2.2.1 ProblemFormulation . . . . . . . . . . . . . . . . . . . . 28
2.2.2 Infinite-HorizonOptimalStateFeedbackControlviaDHP . 30
2.2.3 Simulations . . . . . . . . . . . . . . . . . . . . . . . . . 44
2.3 Infinite-HorizonOptimalStateFeedbackControlBasedonGDHP 52
2.3.1 ProblemFormulation . . . . . . . . . . . . . . . . . . . . 52
2.3.2 Infinite-HorizonOptimalStateFeedbackControlBased
onGDHP . . . . . . . . . . . . . . . . . . . . . . . . . . 54
2.3.3 Simulations . . . . . . . . . . . . . . . . . . . . . . . . . 67
2.4 Infinite-HorizonOptimalStateFeedbackControlBasedonGHJB
Algorithm . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71
2.4.1 ProblemFormulation . . . . . . . . . . . . . . . . . . . . 71
2.4.2 ConstrainedOptimalControlBasedonGHJBEquation . . 73
2.4.3 Simulations . . . . . . . . . . . . . . . . . . . . . . . . . 78
2.5 Finite-HorizonOptimalStateFeedbackControlBasedonHDP . . 80
2.5.1 ProblemFormulation . . . . . . . . . . . . . . . . . . . . 82
2.5.2 Finite-HorizonOptimalStateFeedbackControlBasedon
HDP . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84
2.5.3 Simulations . . . . . . . . . . . . . . . . . . . . . . . . . 102
xi