基于区域扩张策略的势场强化学习算法路径规划研究

未知连续环境状态下的Q学习路径规划算法在执行对环境的试错时收敛速度慢,容易陷入局部,不利于对真实未知环境的探索,为解决此问题,针对Q学习路径规划问题提出一种基于Metropolis准则的区域扩张策略的势场强化学习算法。算法为环境提供势场先验知识初始化状态信息,消除初始时刻的盲目性,提高学习效率,同时引入基于Metropolis准则的区域扩张陷阱区域剔除探索,剔除陷阱障碍物环境的凹形区域。通过MATLAB对多种环境的仿真实验,验证了算法有效性。

Study on Path Planning of Potential Field Reinforcement Learning Algorithm Based on Regional Expansion Strategy

Q-learning path planning algorithm in unknown continuous environment is slow in convergence and easy to fall into local area when executing trial and error on the environment,which is not conducive to the exploration of real unknown environment.To solve the problem,a potential field reinforcement learning algorithm based on Metropolis criterion is proposed for Q-learning path planning.The algorithm provides the environment with the prior knowledge of potential field to initialize the state information,eliminates the blindness at the initial moment and improves the learning efficiency.At the same time,it introduces the exploration of eliminating the trap area of regional expansion based on Metropolis criterion to eliminate the concave area of the trap obstacle environment.The effectiveness of the algorithm is verified by MATLAB simulation experiments in various environments.