机器人 外文翻译 外文文献 英文文献 采用模糊逻辑控制使自主机器人避障设计.docx

机器人 外文翻译 外文文献 英文文献 采用模糊逻辑控制使自主机器人避障设计.docx

《机器人 外文翻译 外文文献 英文文献 采用模糊逻辑控制使自主机器人避障设计.docx》由会员分享，可在线阅读，更多相关《机器人 外文翻译 外文文献 英文文献 采用模糊逻辑控制使自主机器人避障设计.docx（14页珍藏版）》请在冰点文库上搜索。

机器人外文翻译外文文献英文文献采用模糊逻辑控制使自主机器人避障设计

Autonomousrobotobstacleavoidanceusingafuzzylogiccontrolscheme

WilliamMartin

SubmittedonDecember4,2009

CS311-FinalProject

1.INTRODUCTION

Oneoftheconsiderablehurdlestoovercome,whentryingtodescribeareal-worldcontrolschemewithfirst-orderlogic,isthestrongambiguityfoundinbothsemanticsandevaluations.Althoughoneoptionistoutilizeprobabilitytheoryinordertocomeupwithamorerealisticmodel,thisstillreliesonobtaininginformationaboutanagent'senvironmentwithsomeamountofprecision.However,fuzzylogicallowsanagenttoexploitinexactnessinitscollecteddatabyallowingforaleveloftolerance.Thiscanbeespeciallyimportantwhenhighprecisionoraccuracyinameasurementisquitecostly.Forexample,ultrasonicandinfraredrangesensorsallowforfastandcosteffectivedistancemeasurementswithvaryinguncertainty.Theproposedapplicationsforfuzzylogicrangefromcontrollingrobotichandswithsixdegreesoffreedom1tofilteringnoisefromadigitalsignal.2Duetoitseasyimplementation,fuzzylogiccontrolhasbeenpopularforindustrialapplicationswhenadvanceddifferentialequationsbecomeeithercomputationallyexpensiveoroffernoknownsolution.Thisprojectisanattempttotakeadvantageofthesefuzzylogicsimplificationsinordertoimplementsimpleobstacleavoidanceforamobilerobot.

2.PHYSICALROBOTIMPLEMENTATION

2.1.Chassisandsensors

Theroboticvehicle'schassiswasconstructedfromanExcaliburEI-MSD2003remotecontroltoytank.Thedevicewasstrippedofallelectronics,gears,andextraneouspartsinordertoworkwithjusttheemptycaseandtwoDCmotorsforthetanktreads.However,thisleftasomewhatunevensurfacetoworkon,sohigh-densitypolyethylene（HDPE）rodswereusedtofillinemptyspaces.SinceHDPEhasaratherlowsurfaceenergy,whichisnotidealforbondingwithothermaterials,apropanetorchwasusedtoraisesurfacetemperatureandimprovebondingwithanepoxyadhesive.

ThreeSharpGP2D12infraredsensors,whichhavearangeof10to80cm,wereusedfordistancemeasurements.Inordertomounttheseappropriately,a2.5by15cmpieceofaluminumwasbentintothreeevenpiecesat135degreeangles.ThisallowsfortheIRsensorstotakethreedifferentmeasurementsat45degreeangles（right,middle,andleftdistances）.ThissensormountwasthenattachedtoanHDPErodwithmountingtapeandtherodwasgluedtothetankbasewithepoxy.Sincetheminimumdistancethatcanbereliablymeasuredwiththesesensorsis10cm,thesensorswereplacedabout9cmfromthefrontofthevehicle.Thisallowedmeasurementstobetakenveryclosetothefrontoftherobot.

2.2.Electronics

Inordertocontrolthespeedofeachmotor,pulse-widthmodulation（PWM）wasusedtodrivetwoL2722opampsinopenloopmode（Fig.1）.ThehighinputresistanceoftheseICsallowforthemotorstobepoweredwithverylittlepowerdrawfromthePWMcircuitry.Inordertoisolatethemotor'spowersupplyfromtherestoftheelectronics,a9.6VNiCadbatterywasusedseparatelyfromastandard9Vthatdemandontheopampsledtoasmallamountofoverheatingduringcontinuousoperation.Thiswasremediedbyaddingsmallheatsinksandafantotheforciblydisperseheat.

Fig.1.ThecontrolcircuitusedfordrivingeachDCmotor.NotethatthePWMsignalwasbetween0and5V.

2.3.Microcontroller

ComputationwashandledbyanArduinoDuemilanoveboardwithanATmega328microcontroller.Theboardhaslowpowerrequirementsandmodifications.Inaddition,ithasalargenumberofprototypingofthecontrolcircuitandbasedontheWiringlanguage.Thisboardprovidedaneasyandlow-costplatformtobuildtherobotaround.

3.FUZZYCONTROLSCHEMEFOR

Inordertoapplyfuzzylogictotherobottointerpretmeasureddistances.Whilethefinalalgorithmdependedcriticallyonthegeometryoftherobotitselfandhowitoperates,somebasicguidelineswerefollowed.Similarresearchprojectsprovidedbothsimulationresultsandideasforimplementingfuzzycontrol.3,4,5

3.1.Membershipfunctions

Threesetsofmembershipfunctionswerecreatedtoexpressdegreesofmembershipfordistances,translationalspeeds,androtationalspeeds.Thismadeforatotaloftwoinputmembershipfunctionsandeightoutputmembershipfunctions（Fig.2）.Triangleandtrapezoidalfunctionswereusedexclusivelysincetheyarequicktocomputeandeasytomodify.Keepingcomputationtimetoaminimumwasessentialsothatmanysetsofdatacouldbeanalyzedeverysecond（approximatelyoneevery40milliseconds）.ThedistancemembershipfunctionsallowedthedistancesfromtheIRsensorstobequickly"fuzzified,"whiletheeightspeedmembershipfunctionsconvertedfuzzyvaluesbackintocrispvalues.

3.2.Rulebase

Oncetheinputdatawasfuzzified,theeightdefinedfuzzylogicrules（TableI）wereexecutedinordertoassignfuzzyvaluesfortranslationalspeedandrotation.Thisresultedinmultiplevaluesfortheeachofthefuzzyoutputcomponents.Itwasthennecessarytotakethemaximumofthesevaluesasthefuzzyvalueforeachcomponent.Finally,thesefuzzyoutputvalueswere"defuzzified"usingthemax-producttechniqueandtheresultwasusedtoupdateeachofthemotorspeeds.

（a）

（b）

（c）

Fig.2.Themembershipfunctionsusedfor（a）distance,（b）translationspeed,and（c）rotationalspeed.Thesefunctionswereadaptedfromsimilarworkdoneinreference3.

4.RESULTS

Thefuzzycontrolschemeallowedfortherobottoquicklyrespondtoobstaclesitcoulddetectinitsenvironment.Thisallowedittofollowwallsandbendaroundcornersdecentlywithouthittinganyobstacles.However,sincetheIRsensors'measurementsdependedonthegeometryofsurroundingobjects,thereweretimeswhentherobotcouldnotdetectobstacles.Forexample,whentheIRbeamhitasurfacewithobliqueincidence,itwouldreflectawayfromthesensorandnotregisterasanobject.Inaddition,thelimitednumberofrulesusedmayhavelimitedthedynamicsoftherobot'sresponses.Somearticlessuggestasmanyasfortyrules6shouldbeused,whileotherstendtopresentbetweentenandtwenty.Sincethisprojectdidnotexplorecomplexkinematicsorcomputationalsimulationsoftherobot,itisdifficulttodetermine

exactlyhowmanyrulesshouldbeused.However,forthepurposesoftestingfuzzylogicasanavigationalaide,theeightrulesweresufficient.DespitethemanyproblemsthatIRandsimilarultrasonicsensorshavewithreliablyobtainingdistances,therobustnessoffuzzylogicwasfrequentlyabletopreventtherobotfromrunningintoobstacles.

5.CONCLUSION

Thereareseveraleasyimprovementsthatcouldbemadetofutureiterationsofthisprojectinordertoimprovetherobot'sperformance.ThemostdramaticwouldbetoimplementtheIRorultrasonicsensorsonaservosothattheycouldeachscanafull180degrees.However,thistypeofoverhaulmayunderminesomeoffuzzylogic'shelpfulsimplicity.Anotherhelpfultacticwouldbetouseafewtypesofsensorssothatdatacouldbetakenatmultipleranges.TheIRsensorsusedinthisexperimenthadaminimumdistanceof10cm,soanythinginfrontofthiscouldnotbereliablydetected.Similarly,thesensorshadamaximumdistanceof80cmsoitwasdifficulttoreacttoobjectsfaraway.Ultrasonicsensorsdooffersignificantlyincreasedrangesataslightlyincreasedcostandresponsetime.Lastly,definingmoremembershipfunctionscouldhelpimprovetherulebasebycreatingmorefinetunedresponses.However,thiswouldagainincreasethecomplexityofthesystem.

Thus,thisprojecthassuccessfullyimplementedasimplefuzzycontrolschemeforadjustingtheheadingandspeedofamobilerobot.Whileitisdifficulttodeterminewhetherthisisaworthwhileapplicationwithoutheavilyresearchingothermethods,itisquiteapparentthatfuzzylogicaffordsacertainlevelofsimplicityinthedesignofasystem.Furthermore,itisanovelapproachtodealingwithhighlevelsofuncertaintyinreal-worldenvironments.

6.REFERENCES

1Ed.M.Jamshidi,N.Vadiee,andT.Ross,Fuzzylogicandcontrol:

softwareandhardwareapplications,（PrenticeHall:

EnglewoodCliffs,NJ）292-328.

2Ibid,232-261.

3W.L.Xu,S.K.Tso,andY.H.Fung,"Fuzzyreactivecontrolofamobilerobotincorporatingareal/virtualtargetswitchingstrategy,"RoboticsandAutonomousSystems,23（3）,171-186（1998）.

4V.PeriandD.Simon,“Fuzzylogiccontrolforanautonomousrobot,”2005AnnualMeetingoftheNorthAmericanFuzzyInformationProcessingSociety,337-342（2005）.

5A.Martinez,E.Tunstel,andM.Jamshidi,"Fuzzy-logicbasedcollision-avoidanceforamobilerobot,"Robotica,12（6）521–527（1994）.

6W.L.Xu,S.K.Tso,andY.H.Fung,"Fuzzyreactivecontrolofamobilerobotincorporatingareal/virtualtargetswitchingstrategy,"RoboticsandAutonomousSystems,23（3）,171-186（1998）.

采用模糊逻辑控制使1自主机器人避障设计

WilliamMartin威廉马丁

SubmittedonDecember4,2009提交于2009年12月4日

CS311-FinalProjectCS311-最终项目

1.1INTRODUC引言

Oneoftheconsiderablehurdlestoovercome,whentryingtodescribeareal-world其中一个很大的障碍需要克服，当试图用控制逻辑一阶来描述一个真实世界设计在发现在这两个语义evaluations.评价中controlschemewithfirst-orderlogic,isthestrongambiguityfoundinbothsemanticsand设计设计设计是个强大的模糊区。

Althoughoneoptionistoutilizeprobabilitytheoryinordertocomeupwithamo虽然一个方案是利用概率论，以便得到一个更realisticmodel,thisstillreliesonobtaininginformationaboutanagent'senvironmentwithsome现实的模型，这种获得信息的方法的精度仍然依赖于外部环境amountofprecision.。

然而，在其收集databyallowingforaleveloftolerance.数据的公差允许的范围内,模糊逻辑允许利用不精确的间接方法来实现。

在需要高精度oraccuracyinameasurementisquitecostly.测量时它是相当昂贵的。

例如，超声波和红外线传