搜索
单位代码：10293密级：专业学位硕士论文论文题目：RoboCup3D仿真中双足机器人的运动规划与智能决策1210063025郑重虎王保云梁志伟工程硕士申请全日制申请控制工程二○一三年三月学号姓名导师专业学位类别类型专业（领域）论文提交日期TheMotionPlanningandIntelligentDecisionofBipedRobotsinRoboCup3DSimulationEnvironmentThesisSubmittedtoNanjingUniversityofPostsandTelecommunicationsfortheDegreeofMasterofEngineeringByZhengChonghuSupervisor:Prof.WangBaoyunandLecturerLiangZhiweiMarch2013I摘要本文以RoboCup3D仿真机器人足球比赛为平台，对南京邮电大学Apollo3D队足球机器人的个体与整体性能进行了改进和优化。在仿真比赛中，机器人每个周期的位置信息内含有大量的噪声信息，这不仅直接影响到机器人运动的稳定性，而且还影响到机器人对足球的准确定位。为此本文采用了卡尔曼滤波算法来滤除机器人位置信息内含有的噪声，并准确预测出足球在若干个周期之后的位置以提高守门员的防守能力；在动态复杂环境中机器人间常会相互发生碰撞，为此本文研究了在RoboCup3D仿真环境中机器人对运动障碍物的动态避碰问题。针对最短切线避障算法存在的不足之处，提出了改进后的平滑ND（NearnessDiagram）避障算法。算法综合考虑了障碍物速度的动态变化及障碍物活动范围的动态扩展，将对方障碍物看成是一个活动的圆盘，在圆盘的边界处规划出一条避碰路径，并根据对方障碍物的不断变化进行及时追踪；个体机器人在能够很好地完成足球比赛中的自我定位和动态避障行为后，就需要考虑如何将其融入到足球队伍中以提升队伍的整体实力。为此本文采用Q学习和BP（BackPropagation）神经网络相结合的方法使机器人具有一定的学习能力，通过训练学习使机器人在比赛中能够自主调整协作策略以提高控球率，从而达到队伍的整体最优。经过若干场比赛实践证明，上述所提方法的采用大大提高了我校Apollo3D队足球机器人的个体和整体性能，特别是在机器人动态避障、对抗和合作等方面均明显有良好展现，且成绩显著。关键词:RoboCup3D，机器人足球，卡尔曼滤波，动态复杂环境，平滑ND避障算法IIAbstractThispaperpresentedtheimprovementandoptimizationofsoccerrobotsintheRoboCup3DrobotsoccersimulationcompetitionbytheApollo3Dteam.Inthesimulatedenvironment,therobot'ssensorsreceivenoisymeasurementsofpositionalinformationateverycycle.Thesenoisesnotonlyaffectthemotionstabilityofrobots,butalsotheaccuracyofballlocalization.ThispaperemployedtheKalmanfilteralgorithmtoreducethenoise,andimprovetheaccuracyinpositionpredictionofballinnextfewcycles,soastoenhancethegoalie'sdefendingcapability.Inthecomplexdynamicenvironment,therobotscollidewitheachotherfrequently.Therefore,thispaperfirstpresentedashortesttangentalgorithmtomaketherobotdynamicallyavoidcollisionwithmovingobstacles.Inordertohandlethedefectsoftheshortesttangentalgorithm,thispaperproposedanimprovedsmoothND(NearnessDiagram)dynamicalgorithmforavoidingobstacles.Thealgorithmconsideredthedynamicchangingoftheobstacle'svelocityandthedynamicexpandingoftheactiverangeofobstacles,andtookeachobstacleasanactivediscandtrackedtheobstacletimelyaccordingtoitsincessantchange.Asaresult,thealgorithmcanplanacollision-avoidingpathontherimoftheactivedisc.Afteranindividualrobotcanperformwellinself-positioninganddynamicobstacleavoidanceinthesoccergames,itwillneedtolearnhowtoplaytheballasapartoftheteam.ThispapercombinedtheQ-learningalgorithmandtheBP(BackPropagation)neuralnetworktoendowtherobotswithlearningability,whichmakesrobotsadjustcooperationstrategiesindependentlytoimprovetheballcontrollinginthecompetition,soastoachievetheoveralloptimumofperformance.ThecompetitionresultsdemonstratedtheproposedmethodsabovecantremendouslyimproveindividualandoverallperformancesoftheApollo3Dteam,especiallyintherobotdynamicobstacleavoidanceandcoordination.Keywords:RoboCup3D,Robotsoccer,Kalmanfilter,Dynamicandcomplicatedenvironment,SmoothNDobstacleavoidancealgorithmIII目录摘要..........................................................................................................................................................................IAbstract..................................................................................................................................................................II第一章绪论............................................................................................................................................................11.1RoboCup的研究背景................................................................................................................................11.1.1RoboCup的起源.............................................................................................................................11.1.2RoboCup的研究意义.....................................................................................................................21.2研究现状...................................................................................................................................................31.2.1仿人机器人的研究现状................................................................................................................31.2.2RoboCup3D足球机器人的研究现状............................................................................................31.2.3RoboCup3D足球机器人的运动规划与智能决策的研究现状.....................................................41.3主要研究内容...........................................................................................................................................5第二章RoboCup3D足球机器人仿真系统..........................................................................................................72.1Rcssserver3D概述.....................................................................................................................................82.1.1比赛服务器....................................................................................................................................82.1.2比赛平台模型..............................................................................................................................102.1.3ODE系统建模.........