未知环境下一种移动机器人实时最优路径规划方法研究

A Real-Time Optimized Approach to Path Planning for Mobile Robot in Unknown Environment

针对未知环境下移动机器人的安全路径规划问题,提出一种基于改进神经网络和模拟退火算法相结合的方法.神经网络表示机器人的工作空间,通过BP反向算法学习外部环境结构特征和信息表示,进而优化障碍物神经网络的连接权值,利用模拟退火算法搜寻代价函数的负梯度方向,采用组合探测器来减小模拟退火算法搜索区域和应用后退策略及设置虚拟目标点的方法处理局部路径规划中出现的陷阱问题.仿真验证此方法有效性和正确性.

For safe path planning of mobile robot in unknown environment,a method is proposed based on improved neural network and simulated annealing algorithm.Neural network is built to describe the working space of the mobile robot,which connection weights are optimized by the back propagation（BP） learning algorithm to study the structural features and information representation of the environment.Simulated annealing（SA） algorithm by using the combination of detectors to reduce the search area is adopted to get the best negative gradient direction of cost function.A strategy of back strategy and ＂virtual target＂ is introduced to deal with the problem of local minimum,which often occurs in local path planning.The result of the simulation experiment proves the effectiveness and feasibility of the proposed approach.