機(jī)器人外文文獻(xiàn)

1、Keywords:forwardkinematics;sensornetwork;sensorfusion;FPGA;industrialrobot1.IntroductionFlexiblemanipulatorrobotshavewideindustrialapplications,withhandlingandmanufacturingoperationsbeingsomeofthemostcommon1-3.High-precisionandhigh-accuracyinrobotoperationsrequirethestudyofrobotkinematics,dynamicsan

2、dcontrol4.Theaimofforwardkinematicsistocomputethepositionandorientationoftherobotendeffectorasafunctionoftheangularpositionofeachjoint1.Theonlineestimationoftheforwardkinematicscancontributetoimprovethecontrollerperformancebyconsideringthejointsmotioncollectively.Therefore,theprecisionandaccuracyofs

3、uchinformationisessentialtothecontrollerinordertoincreaseitsperformanceinrealroboticoperations.Commerciallyavailablemotioncontrollersuseasinglesensorforeachjointtoestimatetherobotsangularposition;themostcommonsensoristheopticalencoder5-11,whichprovidesahigh-resolutionfeedbacktothecontroller.However,

4、itonlygivesinformationontheservomotorpositionandanydeformationscausedbyjointflexibilitiescannotbemonitored6,12,decreasingtherobotsaccuracy.Thisproblemismoreevidentinopen-chainrobots.Moreover,theprovidedinformationisrelative,whichmeansthatitisimpossibletoestimatetheinitialpositionoftherobot.Anotherse

5、nsorthatiswidelyusedintheestimationoftheangularpositionoftherobotjointsisthegyroscope;itprovidesameasurementofangularrateofchange,requiringtheaccumulatedsumovertimetoestimatetheangularposition.Despitethefactthattheycandetectsomenonlinearitiesthatcannotbeestimatedwithencoders,thequantized,noisysignal

6、causesaccumulatederrorswhenangularpositionisrequired13-15.Furthermore,theestimatedangularpositionisrelative,whichdoesnotpermitonetoknowtheinitialangularpositionoftherobotjoints.Agoodsensorthatprovidesanabsolutemeasurementistheaccelerometeranditcanbeusedtoestimatetherobotangularposition5,16-20.Nevert

7、heless,thesignalobtainedisnoisyandcontainsmuchinformationthatneedspreprocessingbeforebeingused21.Twomainissuesneedtobesolvedwhentherobotforwardkinematicsisrequired:theproblemsofusingasinglesensortoestimatetheangularpositionofthejointsandtheonlineestimationoftheforwardkinematics.Inthisperspective,sen

8、sorfusiontechniquesimprovetheaccuracyofthemonitoredvariables,butattheexpenseofhigh-computationalloads22,whichcomplicatetheonlineestimationoftheforwardkinematics.Someworkscombinesensorfusiontechniquesandforwardkinematicsestimation.Forexample,in7accelerometerandencodersignalsarefusedusingadisturbanceo

9、bservertocompensatesomenonlinearitiesandaresetstateestimatorforpositionestimation;experimentationisperformedonalinearmotorpositioningsystem,whichrequiresnoforwardkinematicsestimation.Anotherworkthatfusesencoderandaccelerometersignalsispresentedin16,wheretheforwardkinematicsoftwolinksina6-DOFrobotisc

10、alculated;differentversionsofanextendedKalmanfilterareusedforsensorfusion.However,theefficacyoftheproposedalgorithmisdemonstratedoffline.Otherworksattempttofusemorethantwodifferentsensors.In12thefusionofencoder,accelerometerandinterferometerthroughaKalmanfilterispresentedtoestimatethepositionofapara

11、llelkinematicmachine,buttheanalysisislimitedtoone-axismovement.In6camera,accelerometerandgyroscopesensorsarecombinedthroughakinematicKalmanfilterforpositionestimationofaplanartwo-linkrobottofacilitatetheforwardkinematicsestimation.In23,24ahardware-softwarearchitectureforsensornetworkfusioninindustri

12、alrobotsisproposed.MultiplePCstoprocessallthedatacollectedfromthesensorsandtocontroltherobotareused.However,theyusethesensorfusiontoestimatetherobotcontactforceandtheforwardkinematicsareestimatedonlyfromtheencodersignals.Reportedworksnotethelimitationsofusingasinglesensortoestimaterobotsforwardkinem

13、atics.Besides,forwardkinematicsislimitedtoacoupleofjointsduetotheequationscomplexity.Therefore,theonlineforwardkinematicsestimationproblemformulti-axisrobotsstillrequiresadditionalefforts.Duetothis,adedicatedprocessorcapableoffilteringandfusingtheinformationofseveralsensorswouldbedesirable.Also,mult

14、i-axisforwardkinematicsestimationinanonlinefashionwouldbeadvantageous.Thecontributionofthisworkisperformedoftwostages:theimprovementofthesensingmethodofconventionalmotioncontrollersthroughproposalofanencoder-accelerometer-basedfusedsmartsensornetwork.Furthermore,weproposeasmartprocessorcapableofproc

15、essingallthesensedencoder-accelerometersignalssoastoobtainonlineforwardkinematicsestimationofeachjointofasix-degree-of-freedom(6-DOF)industrialrobot.Thesmartprocessorisdesignedusingfield-programmablegatearrays(FPGA)owingtotheirparallelcomputationcapabilitiesandreconfigurability.Itisdesignedthroughth

16、ecombinationofseveralmethodsandtechniquestoachieveonlineoperation.Thesmartprocessorisabletofilterandtofusetheinformationofthesensornetwork,whichcontainstwoprimarysensors:anopticalencoder,anda3-axisaccelerometer;andthentoobtaintherobotforwardkinematicsforeachjointinanonlinefashion.Thesensornetworkisc

17、omposedofsixencodersandfour3-axisaccelerometersmountedontherobot.Anexperimentalsetupwascarriedoutona6-DOFPUMA(ProgrammableUniversalManipulationArm)robot,demonstratingtheperformanceoftheproposedfusedsmartsensornetworkthroughthemonitoringofthreereal-operation-oriented3Dtrajectories.Moreover,additional

18、experimentswerecarriedoutwherebytheforwardkinematicsobtainedwiththeproposedmethodologyiscomparedagainsttheobtainedthroughtheconventionalmethodofusingencoders.Animprovementinthemeasurementaccuracyisfoundwhentheproposedmethodologyisused.2.MethodologyTheuseofaccelerometerson6-DOFPUMArobotsrequiresplaci

19、ngthemadequatelyinspecificpositions.Thecombinationofaccelerometersandencodersmakeupthesensornetworkthatneedstobeprocessedinordertoestimatetheangularpositionofeachjointandtheforwardkinematicsaccurately.Inthissection,theplacementofthesensornetworkonthePUMArobotispresented.Then,theFPGA-basedforwardkine

20、maticssmartprocessorisclearlydescribed.2.1.SensorNetworkAsensornetworkisanarrayofdiversetypesofsensorstomonitorseveralvariables25,26;inthiscasetheangularpositionofthejointflexibilitiesoftherobot.ThesensornetworkarrangedontherobotispresentedinFigure1.Figure1(a)depictsthepositionoftheaccelerometersont

21、herobot.xiA,yiAandziAarethemeasurementsofeachaxisfromtheaccelerometeriA.Figure1(b)isaschematicoftherobotincludingitslinkparameters.irepresentstheangularpositionofjointi.iaandidrepresentstherobotphysicaldimensions.Also,thelocalizationofencodersiEisshown.Figure1.Sensornetworklocationontherobot.(a)Acce

22、lerometerplacement.(b)Schematicshowingtheparametersthatdescribetherobot.Forthiswork,theforwardkinematicsestimationconsistsontheestimationofthejointangularposition(i),thespatialpositionofeachjoint(iiiZYX);andtheroll(i),pitch(i),andyaw(i)angles1.Suchanglesrepresenttherotationalongthe000XYZaxes,respectively,requiredtoobtaintheorientationofeachjoint.2.1.1.AngularJointPositionThejointangularpositioniscalculatedwithbothencoderandaccelerometersensors.Afterwards,theobtainedinformationisfus