基于回声状态网络的进化机器人路径规划方法

A Method of Path Planning for Evolutionary Robots Based on Echo State Network

在移动机器人路径规划问题的研究中,针对进化机器人在不同环境下需要重新进化与学习的问题,为了使进化机器人准确定位,提出了一种基于(μ+λ)-ES进化策略的回声状态网络路径规划算法,利用回声状态网络构建移动机器人传感器输入和执行器输出之间的映射关系,利用(μ+λ)-ES进化策略对回声状态网络进行无监督学习,机器人利用进化获得的神经网络控制器进行路径规划。仿真结果表明,根据回声状态网络的路径规划方法对于机器人动态未知环境具有较好的实时性和适应性,达到准确定位目标。

For the re-evolution of the mobile robot behavior in unknown environments,an algorithm of path planning was presented based on recurrent neural network for evolutionary robots.The mapping relation was constructed between input of sensors and output of actuators based on echo state network.The（μ＋λ）-Evolution Strategy is used to optimize the output weights of echo state network.The robot can generate the optimal trajectory under dynamic environments via the neural network controller.The experimental results indicate that the proposed approach has high adaptability and real-time ability under dynamic environments.