时变流场环境中机器人跟踪气味烟羽方法

3510Vol.35,No.10200910ACTAAUTOMATICASINICAOctober,200911111..,,.,.,.,,,TP24TracingOdorPlumebyRobotinTime-variantFlow-¯eldEnvironmentsLIJi-Gong1MENGQing-Hao1LIFei1JIANGPing1ZENGMing1AbstractThereliableplumetracingisacriticalissuefortheimplementationofodorsourcelocalization.Thispaperaddressestheproblemoftracingplumebyamobilerobotinrealtime-variant°ow-¯eldenvironments.Anestimatimationofthemostlikelypathtakenbytheodorpatchistobedonewhenevertheodorpatchisdetectedbythegassensormountedonthemobilerobot.Subsequently,asearchingpathiscalculatedandfollowedbythemobilerobottotracetheodorplumewithconsiderationofthelikelyodorpathestimatedandthe°owdirectionatcurrenttime.Therelativechangeofodorconcentrationisadoptedastheconcentration¯eldistime-variantandmighthaveacommonfoundationalconcentration.Theexperimentresultintheindoortime-variantair°ow-¯eldenvironmenthasshowthatthemobilerobotcanbee®ectivelyguidedinrealtimewiththeproposedmethodintracingplumeandreachingtheodorsource.KeywordsMobilerobot,activeolfaction,time-variant°ow¯eld,plumetracing.,90[1].[2][3],[4].\,.()[5]./.Hayes[6]2008-08-202009-03-20ReceivedAugust20,2008;inrevisedformMarch20,2009(863)(2007AA04Z219),(60875053,60802051)SupportedbyNationalHighTechnologyResearchandDe-velopmentProgramofChina(863Program)(2007AA04Z219),NationalNaturalScienceFoundationofChina(60875053,60802051),andtheProgramforNewCenturyExcellentTalentsinUniversity(NCET)1.3000721.SchoolofElectricalEngineeringandAutomation,TianjinUniversity,Tianjin300072DOI:10.3724/SP.J.1004.2009.01327,3,(Plume¯nding)(Plumetraver-sal)(Odorsourcedeclaration).Li[7]4:(Findingaplume)(Tracingtheplume)(Reac-quiringtheplume)(Declaringthesourcelocation).,.,.,,[7];,,,,,..,;,132835,,./:,.,,:(Spiral-surge)[6][8]Z(Zigzag)[7;9].,;,;Z,.,(Chemotaxis)[9](Anemotaxis)[10]..(Fluxotaxis)[11](Infotaxis)[12]..,,.,(),(Clover-leaftrajectory)[7].,,[7]..(/),.,\,,.(),;(/),.\,.(),.Kowadlo[13](),,.Farrell[14](HiddenMarkovmodel,HMM),.Pang[15],Bayesian..,,,[16]().,,.1¿,tk0()c(tk0¡¿)»=c(tk0¡h¢t)=c(tk0¡h)=c(tk)(1)tk(tk0).,h=int(¿=¢t),k0k,k0=k+h;¢t;int(¤)¤.t0.[7;14¡15],.,,().,,.,¹c(tk)=(¸¹c(tk¡1)+(1¡¸)c(tk);k¸1c(tk);k=0(2)¸2[0;1],¸=0:5,,c0(tk)=(c(tk)¡¹c(tk¡1);c(tk)¹c(tk¡1)0;(3)c0(tk)0,,.2tiLR(ti)/,xy10:1329(ux(LR(ti));uy(LR(ti))),/f(ux(LR(ti));uy(LR(ti)))gk0i=f.,tf/,f=max(0;k0¡int(tthr=¢t)),tthr;µ(ti)=tan¡1(uy(LR(ti))=ux(LR(ti))).,[7]Z.tk(tk0¡h,tk0)LR(tk)=[xj;yj]T,.[16],,tl2[tf;tk]DW(tl;tk),Lmax(tl)=[xj¡vx(tl;tk);yj¡vy(tl;tk)]T,,vx(tl;tk)vy(tl;tk)tltkxy,vx(tl;tk)¼Pki=lux(LR(ti))¢t,vy(tl;tk)¼Pki=luy(LR(ti))¢t.Lmax(tl)tl,[tf;tk]fLmax(tl);l=k;k¡1;¢¢¢;fg.,,,.,,,,,1.(),:fLo®(tl);l=k;k¡1;¢¢¢;fg,,Lo®(tl)=Lmax(tl)+(dell(tl)+dbas)¢cosµ(tk0)sinµ(tk0)#(4),(dell(tl)+dbas)¢[cosµ(tk0);sinµ(tk0)]TLo®(tl)Lmax(tl),2.2,dell(tl)=p2(tk¡tl)K(tl)¢q(¾xcosÃ)2+(¾ysinÃ)2(5a)Ã=tan¡1·¾x¾y¢tanµ(tk0)¸(5b),K(tl)=ln(A=(2¼(tk¡tl)¾x¾y))[16],A;(¾2x;¾2y)xy,f(ux(LR(ti));uy(LR(ti))gki=f,¾2x=var(ux(LR(ti));i=f;¢¢¢;k),¾2y=var(uy(LR(ti));i=f;¢¢¢;k);ÃDW(tl;tk)X;dbas;tk0(tk0¸tk);µ(tk0)tk0.1Fig.1Tracingplumebasedontheestimatedodorpath2Fig.2Thegenerationofdeviation-path-point,tk0.,µ(tk0),LR(tk0),,.,,(),,,¹µ(tk0)=tan¡1(vy(tf;tk0)=vx(tf;tk0))133035,,,1.Casting.,,,,.3,,.3Fig.3The°owchartofodorsourcesearch,:SL=fSLf;SLe;SLbg,.,SL,SLfSLb,SLe,:SLf=©Lo®(tl)¯¯jang(Lo®(tl)¡LR(tk0))¡µ(tk0)j¼=2;l=k;k¡1;¢¢¢;fg(6a)SLf=©Lo®(tl)¯¯jang(Lo®(tl)¡LR(tk0))¡µ(tk0)j·¼=2;l=f;f+1;¢¢¢;kg(6b)SLe=©L(i)e;i=0;1;¢¢¢;8ª(6c),ang(¤)¤,ang(¤)2(¡¼;¼];L(i)e=Lmax(tf)¡sin¹µ(tk0)cos¹µ(tk0)¡cos¹µ(tk0)sin¹µ(tk0)#¢Rlsin(clk¢i¼=4)Rs(1¡cos(clk¢i¼=4))#,RlRs,clk,Lmax(tf)(tk00)¢¹µ(tk00),clk=(1;¢¹µ(tk00)¸0¡1;,¢¹µ(tk00)=¹µ(tk00)¡¹µ(tk00¡1),¢¹µ(tk00)2(¡¼;¼].,.SL(),,(6)li():1),;2).3(a)(b)(a)Themobilerobotand(b)Theconstructionoftheonboardsensorstheolfactorysystem4Fig.4Experimentplatform10m£10m,.,0.1m.4(a),/10:1331.;,,;,,CCD,,4(b);.,.,DSP;PC(CPU:1.5GHz,RAM:256MB).¢t0.5s.¿¼5s,.tthr=15s,/31(/31/||/.31,;31,,).5s,/5s,.,,0.15m/s.dbas=0:1m((4)),Rl=3Rs=0:6m((6c)).,0.5s.·0:6m,,.5,S5.0m,199.5s.,5(a)5(b).0.01»1.05m/s,0±»334±,5.0»13.2ppm,6(/).,5,,[16],.,(),.,,0.6m,VFF(Virtualforce¯eld)(),0.6m.(a)t=98:0s(b)t=199:5s5Fig.5Theprocessofodorsourcesearch1332355(a)t=98:0s,246±,230±,,t=89:5st=98:0s.,BA(S),A,SB;BA,B(A),.,,5(b).,,,.(cm/s,±,£20¡1ppm)6/Fig.6Sensordataofconcentrationandwindduringtheodorsourcesearch4.,,.,,;,,.References1IshidaH,SuetsuguK,NakamotoT,MoriizumiT.Studyofautonomousmobilesensingsystemforlocalizationofodorsourceusinggassensorsandanemometricsensors.SensorsandActuatorsA:Physical,1994,45(2):153¡1572RussellRA.Trackingchemicalplumesinconstrainedenvi-ronments.Robotica,2001,19(4):451¡4583LytridisC,KadarEE,VirkGS.Asystematicapproachtotheproblemofodoursourcelocalization.AutonomousRobots,2006,20(3):261¡2764MengQing-Hao,LiFei.Reviewofactiveolfaction.Robot,2006,28(1):89¡96(,..,2006,28(1):89¡96)5PayneTL,BirchMC.MechanismsinInsectOlfaction.Oxford:OxfordUniversityPress,1986.27¡386HayesAT,MartinoliA,GoodmanRM.Distributedodorsourcelocalization.IEEESensors,2002,2(3):260¡2717LiW,FarrellJA,PangS,ArrietaRM.Moth-inspiredchem-icalplumetracingonanautonomousunderwatervehicle