基于实际函数的机器人模糊神经网络路径规划  

FNN path planning of robot based on actual functions

作　　者：张寒松[1,2] 贾瑞清[2] 

机构地区：[1]中冶京诚工程技术有限公司炼钢工程技术所,北京100176 [2]中国矿业大学(北京)机电与信息工程学院

出　　处：《矿山机械》2010年第14期44-48,共5页Mining & Processing Equipment

基　　金：国家自然科学基金资助项目(50475094)

摘　　要：介绍了一种机器人在不确定环境下,通过探障传感器,在探测过程中避障并最终达到目标的在线路径规划方法。它把所探障碍物当作运动物考虑,使该路径规划对于静止和运动障碍物都能恰当处理。引入有实际物理意义的模糊隶属函数来计算避碰隶属度,采用模糊神经网络来实现模糊控制法则,以避免模糊规则的死点从而完成避碰路径规划。采用有实际物理意义的误差函数对神经网络的权值进行调整,再通过避碰路径和寻的路径的融合使机器人最终避开障碍物而达到目标。仿真试验验证了该方法的有效性和实用性。This paper introduces a kind of path planning method for robot to evade obstacles and reach the objective with the aid of obstacle detecting sensor in uncertain surroundings.All the detected obstacles were considered as mobile ones,so static and mobile obstacles could be handled appropriately by the path planning.The fuzzy membership function containing actual physical content was introduced to calculate the membership value of evasion,and the fuzzy neural network（FNN） was used to realize fuzzy control rules to evade the blind points of fuzzy rules so as to achieve evasion path planning.Moreover,error function containing actual physical content was applied to adjust the weights of FNN,and integration of evasive paths and objective seeking paths was to guide the robot to evade obstacles and reach the objective.The method proves available and practical by simulation experiment.

关 键 词：机器人 路径规划 隶属函数 误差函数 模糊神经网络 

分 类 号：TP242[自动化与计算机技术—检测技术与自动化装置]