摘要
介绍了输电线路除冰机器人的研究现状。针对其工作环境恶劣、不确定因素多的特点,提出了基于马尔可夫决策的行为控制器的设计方法。该方法首先定义了输电线除冰机器人的马尔可夫模型,然后给出了相应的最优方针搜索策略,还给控制器添加了概率调节机制,以达到执行效果反馈的目的。为解决机器人作业中的突发情况,在该方法中还引入了人工辅助和行为学习网络。给出了控制器判断机器人需要申请人工辅助的算法。在人工辅助过程中,通过行为学习网络,机器人可以学习人工辅助中操作员的动作,并利用现有行为完成更加复杂的新任务。实验表明该方法不但可以根据机器人的状态规划出最优动作,还可以在线更新控制策略,具有很强的灵活性和鲁棒性。
The present situation of the research of transmission line de-icing robot is introduced. A behavior controller based on Markov Decision Process（MDP） is presented for robot＇s poor working environment and uncertainty. First the MDP model of the transmission line de-icing robot is defined, and then the search strategy of optimal policy is given, finally the mechanism of probability adjustment is added for the feedback of the result of implementation. Artificial auxiliary and learning behavior networks is introduced for the emergency situations in the operation of robot. The algorithm of determining the need of artificial auxiliary for robot is given. In the process of artificial auxiliary, robot can learn the action of operator in artificial auxiliary with learning behavior network, and complete more complex new tasks by exist behaviors. The experiment results show that controller has a strong flexibility and robustness, it not only give the optimal action by the state of robot, but also update the control strategy online.
出处
《控制工程》
CSCD
北大核心
2011年第3期434-438,共5页
Control Engineering of China
基金
国家863项目(2008AA04Z214)
国家支撑计划项目(2008BAF36B01)
国家自然科学基金重点项目(60835004)
