欢迎来到天天文库
浏览记录
ID:39402825
大小:1011.00 KB
页数:74页
时间:2019-07-02
《基于神经网络的智能移动机器人导航算法研究》由会员上传分享,免费在线阅读,更多相关内容在行业资料-天天文库。
1、浙江工业大学硕士学位论文基于神经网络的智能移动机器人导航算法研究姓名:李佳鹤申请学位级别:硕士专业:控制理论与控制工程指导教师:姚明海20091110浙江工业大学硕士学位论文的学习,提出一种ARTQL神经网络模型,并将它应用于智能机器人导航控制中。ARTQL神经网络将Q学习算法的“试错法”行为探索策略和ART2神经网络的自适应竞争学习机制相结合,利用网络中的长期记忆LTM权向量来存储机器人探索过的所有状态空间模式特征以及相应的模式.行为估计值,使其无需任何先验知识即可在完全未知环境中进行行为决策和状态空间模式分类的两层在线学习,通过与环境的交互,不断总结和积累不同环境特
2、征下的导航经验,逐步改进控制策略,提高学习精度。关键词:导航控制,神经网络,模糊神经网络,遗传进化神经网络,ARTQL神经网络浙江工业大学硕士学位论文RESEARCHoNNEURALNETWORKBASEDINTELLIGENTMoBILERoBOTNAVIGATIoNALGOIUTHMABSTRACTMobilerobotisanimportantdevelopingdirectioninthefieldofrobotics,itiswidelyusedinindustry,agriculture,military,transportation,services,er
3、e.Navigationcontroltechnologyisthecoreintheseresearchareas,whichisalsothekeytechnologyformobilerobottoachieveintelligentizationandautonomation.Theartificialneuralnetworkhasbeenwidelyusedinthefieldofmobilerobotnavigationcontrolbecauseofitsstronglearningabilityandthenon—linearmappingcapabi
4、lity.Thisthesisdiscussesandresearchestheneuralnetworktechnologyanditsapplicationinmobilerobotnavigationcontroldeeply,andforthecharacteristicsandresearchofmobilerobotnavigationproblem,itfindthewaytoimprovethemobilerobotintelligentleveloftheself-organization,self-learning,adaptivecapabilit
5、yandsolveautonomousnavigationproblemofthemobilerobotinthecomplexuncertainenvironmentsandunknownvariableenvironments,byusingthetechnologyofneuralnetworkthatcombinationwithotherintelligentalgorithmssuchasreinforcementlearning,etc.Themainworksandachievementsofthisthesisisasfollows:1.Anaviga
6、tioncontrolmethodbasedonfuzzyneuralnetworkhasbeenresearchedfortheproblemofmobilerobotreal-timenavigationcontrolintheuncertainenvironment.Thequalitativeknowledgeisintroducedintheneuralnetworkframework.Afuzzylayerisinsertedbetweeninputlayerandoutputlayerintheconventionalneuralnetworktomake
7、theneuralnetworkconstructed、析mthefuzzyrules.whichcanretainthelearningmechanismofneuralnetworkswhilemakingtheweightsinthenetworkhaveaclearphysicalmeaning.Thenetworkcombinedwithself-learningabilityofneuralnetworksandtheintellectualstructureoffuzzylogichasagreaterpertinencet
此文档下载收益归作者所有