机器人视觉伺服控制系统研究

西安理工大学硕士学位论文机器人视觉伺服控制系统研究姓名：辛菁申请学位级别：硕士专业：控制理论与控制工程指导教师：刘丁;刘涵20030301I::::2003.3GAPD2PDPDPDPDPDPDMOTOMANMOTOMANMOTOMANCCDPCIIlookandmoveMOTOMANGAMOTOMANAbstractIIIRESEARCHONROBOTVISUALSERVOCONTROLSYSTEMSpeciality:ControlTheory&ControlEngineeringCandidate:JingXin(Signature:)Supervisor:Prof.DingLiu(Signature:)Lect.HanLiu(Signature:)DateofOralExamination:March,2003ABSTRACTRobotvisualservocontrolisafundamentalresearchsubjectwithveryimportanttheoreticalresearchsenseandindustryapplicationprospect.Firstly,AmethodforobjectrecognitioninacomplexnoisyenvironmentbasedonGApatternmatchingispresentedinthisthesis,combiningwithdistancetransformation.Simulationsshowthatthismethodcanquicklyandaccuratelyrecognizetheobjectininputimageandhasafewstrongrobustnessagainstnoiseintheinputimage.Secondly,ThesimulationsfortwokindsofcontrolalgorithmsbasedonPD,namely,directPDcontrolandPDcontrolwithfeedforwardcompensation,havebeencompletedinthisthesis.SimulationsshowthatbothcontrolalgorithmshavegoodcontrolperformanceintrackingaknowntrajectoryandtheeffectofdynamiccompensationofPDcontrolwithfeedforwardcompensationisobvious.ButdirectPDcontrolhasbetterrobustnessagainstmodelingerroranduncertaindisturbancecomparedwithPDcontrolwithfeedforwardcompensation.Finally,theforwardkinematicsmodelofMOTOMANrobotissetupanditsinversekinematicssolutionisobtainedbyusingalgebraicmethodanditerativemethodrespectively;AMOTOMANindustryrobot,aCCDcameraandanimagegrabbercard,alongwithPCAbstractIVhostcomputer,formedeye-in-handcoordinatedvisualservocontrolsystem,whichconstructedahardwareplatformofthetheoreticalresearchandsimulationexperimentforrobotreal-timevisualservocontrol.The3Dvisionlocatingofobjectisdoneusing“hand-eye”stereovisionlocatingalgorithmsbasedonthisexperimentsystem.Experimentresultsshowthatthealgorithmiseffectiveaswellassettlement.Somemeasuresforimprovingthelocatingprecisionarepresentedalso.Arobotvisionlocatingexperimentiscompletedbasedondynamicposition-basedlook-and-movevisualservocontrolstructureandexperimentresultsshowthattherobotvisionsystemhashighlocatingprecision.KEYWORDS:GApatternmatching;MOTOMANrobot;visualservocontrol;eye-in-handcoordinatedvisualservocontrolsystem111.1,Alaincodourey[1]NASAARTRAROTEXESPRITEUREKA863SRISRIVISIONMODULEPhilipsPAPS;200111ASIMOCCDlookandmove[2][3],[4]2,[2]()(CCD)lookandmove1.23[5]HillPark1979[6],()6tx[5]x3D6tcxctx1-12D3D3Dctx6tcx06tx1-1world0,end-effectort6,cameracandtargett46DOF(degreesoffreedom),3D2DOF1980SandersonandWeiss[7]1-21-5(3)[7,8],(2)Espiau[9]lookandmovePBVSIBVSlookandmove[5]1.2.53.[10]____dxcˆxcfFigure1-2Dynamicposition_basedlook_and_movestructure1-2Figure1-3Dynamicimage_basedlook_and_movestructure1-3____dff____dxcˆxcfFigure1-4Position_basedvisualservo(PBVS)structureasperWeiss1-4__dffFigure1-5image_basedvisualservo(PBVS)structureasperWeiss1-56EOLendpointopen-loopECL(endpointclosed-loop)EOL-EOL-ECL-[2,11,12]ECLEOLECLEOLECLECL1.3lookandmoveMOTOMANSV3GAGA72D3DMOTOMANMOTOMANPD2PDPDPDPDPDPDCCDPCMOTOMANSV3”lookandmove”MOTOMANSV3822.1[13~17]199840(2-1)(1)(2)1.(a)(b)2-191G__2-2FFT[18][19,20]2.[19,21,22]2-210(GeneticAlgorithms,GA)GA,,,GA,,GAGA2.2GA2.2.1oφmφS()mFφS()mFφmφFomφφ≠()()omFFφφMθ()ccyx,**cossinsincosiciicixxxMyyyθθθθ−=••+(2-1)()iiyx,()**,iiyx11NnRFb==(2-2)R10≤≤Rbn()()iiiiyxfyxf,,**=()iiyxf,()**,iiyxf,N()θ,,,MyxccR,,,,,,,,GA2-32.2.2(),nPGA2-3GA12[23]GA(1)(2)(3)GA(4)1GA2-42.2.3a.GARYN2-4GA13(distancetransformation)(1L),-1LL0(),xiyiBBL1(),xiyiWWL1000L101d{}max,xixiyiyidWBWB=−−0Ld−2-51L+4L=()NLyxfXFNiii*,)(**∑=(2-3)(a)(b)2-514NiGAGAGAGAGA00GA()fx()Fx()()Fxafxb=⋅+(2-4),ab10.2~0.115Fx=f(x)(2-5)Fx=kxf)((2-6)k2≥kb.()θ,,,MyxccR4(),ccxyθM2.2.4a.128×12819×2134PopSize=80;MaxGen=100;2-6[][][][]1,1281,1280.8,1.2/2,/2ccxyMpipiθ∈∈∈∈−2-7b.X01…010…100…110…199792-6ccxyMθ16()NLyxfXFNiii*,)(**∑=2-8c.1(roulettewheelselection)20-111101110011010210101100101(1101)101100011010210011100101111101111111200110000000()yx,Mθ17Pc=0.9;3Pm=0.1Pm=0.01()d.2-72-82-7(a)2-8(a)128128,2-8(a)2-7(a)(b),(c)(d)(e),(f),+100,2-7(a)96,2-8(a)95,()68sMATLAB(a)(b)(c)(d)18GA2-9GA(d)(e)(f)2-7GA1FitnessValueGenerationtimes(a)(b)(c)(d)2-8GA2(d)(e)(f)FitnessValueGenerationtimes192.3,,,,,NmlφNlθ1NmφNmnφ()11111,,,NNNNNmxyMφθ=GANoφx1Nmlφ+1Nlθ+11Nmφ+1Nmnφ+()1111111111,,,NNNNNmxyMφθ+++++=GAN+1oφxy2-9GA2033.12D3D2D3D[24],,[25]--3.2[26],[27]21,3.2.1-cS,eS,01eecxyzecRPSS=(3-1)ecR33exyzPecRexyzPeSoS,01ooexyzoeRPSS=(3-2)oeR33,,oxyzP,(3-1)(3-2)0101,101ooeeexyzcxyzocoeoeoTecexyzxyzcxyzRPRPSSRRRPPS=+=(3-3)(3-3)cSooecoeoxyzecxyzexyzxyzSRRSRPP=++(3-4)22..0...0.cxcuzcycvzfSfxzkSfSfyzkSuuvv==+==+(3-5)[],TuvcS,f,,,cccxyzSSScS,ukvkXY,0u0vXY11ff−=(3-5)()()100,,TccxyzzuvSSfkuukvvf−=⋅−−(3-6)(3-4)(3-6)()()[]()100100,,[,,1],,0[,,]TooecxyzeczuvoeoexyzxyzToecTeczuvoeoexyzxyzSRRSfkuukvvfRPPRRSfdiagkkuvuvfRPP−−=−−++=+−−++(3-7)[][][]100,,0,,133,,3131ToccoxyzzzxyzoeecuvToeexyzSSKuvSLMPKRRfdiagkkLKuvfMRP−=⋅+⋅++=⋅⋅∈×=−−∈×=⋅∈×(3-8)3.2.2,oeR,oeR,23[]32,ToccoxyzzzxyzSSKuvSLMP×=+++(3-9),,KLM,K32K×,,KLM,oxyzS[],,TczSuvoxyzP[],Tuv