Application of Growing Self-organizing Neural Network to Reinforcement Learning in Obstacle Avoidance

摘要：

This paper focuses on the growing self-organizing neural network used for reinforcement learning in obstacle avoidance of mobile robot.Learning ability is essential for the robot and reinforcement learning is used widely.Q-learning is a widely used reinforcement learning type,but it only used in discrete states and a large memory is needed to store the Q-value.Self-organizing map is a competitive neural network and preserves the topology mapping from the high dimension into a low dimension.Growing self-organizing neural network is developed on the basis of self-organizing map for more complex applications.In order to solve the reinforcement learning problems,growing self-organizing network is applied for quantization of the state space.Experiment results show the efficiency of the method in obstacle avoidance.

关键词： reinforcement learning growing self-organizing neural network obstacle avoidance mobile robot.

作者: Jun-fei Qiao Zhan-jun Hou Xiao-gang Ruan

作者单位: Institute of Artificial Intelligence and Robots,Beijing University of Technology,Beijing,China

会议类型: 国际会议

会议名称: International Conference on Modelling,Identification and Control(模拟、鉴定、控制国际会议)

会议地点: 上海

会议语种:英文

在线出版日期: 2008-06-29（万方平台首次上网日期，不代表论文的发表时间）