强化学习在移动机器人自主导航中的应用  

Application of reinforcement learning in autonomous navigation for mobile robot

作　　者：秦政[1] 丁福光[1] 边信黔[1] 

机构地区：[1]哈尔滨工程大学自动化学院,哈尔滨150001

出　　处：《计算机工程与应用》2007年第18期215-217,228,共4页Computer Engineering and Applications

摘　　要：概述了移动机器人常用的自主导航算法及其优缺点,在此基础上提出了强化学习方法。描述了强化学习算法的原理,并实现了用神经网络解决泛化问题。设计了基于障碍物探测传感器信息的机器人自主导航强化学习方法,给出了学习算法中各要素的数学模型。经仿真验证,算法正确有效,具有良好的收敛性和泛化能力。The merit and shortcomming of common algorithms of autonomous navigation for mobile robot are introduced,based on which the reinforcement learning method is proposed.The principle of the reinforcement learning is described,the generalization problem is solved by neural network.The autonomous navigation for robot based on obstacle detection sensor is designed,the mathematical model for each element of learning algorithm is proposed.The correctness ,reactiveness and the ability of generalization of the algorithm are verified by simulation tests.

关 键 词：强化学习 自主导航 神经网络 移动机器人 

分 类 号：TP24[自动化与计算机技术—检测技术与自动化装置]