基于深度Q学习的移动机器人路径规划

Robot Path Planning Based on Deep Q-Learning

针对传统Q-learning算法在复杂环境下移动机器人路径规划问题中容易产生维数灾难的问题,提出一种改进方法。该方法将深度学习融于Q-learming框架中,以网络输出代替Q值表,解决维数灾难问题。通过构建记忆回放矩阵和双层网络结构打断数据相关性,提高算法收敛性。最后,通过栅格法建立仿真环境建模,在不同复杂程度上的地图上进行仿真实验,对比实验验证了传统Q-learming难以在大状态空间下进行路径规划,深度强化学习能够在复杂状态环境下进行良好的路径规划。

In order to solve the problem that the traditional Q-learning algorithm is prone to dimension disaster in the path planning of mobile robot in complex environment,an improved method is proposed.This method in-tegrates deep learning into the Q-learning framework and replaces the Q-value table with network output to solve the dimensionality disaster problem.In addition,by constructing a memory playback matrix and a two-layer network structure,data correlation is interrupted to improve the convergence of the algorithm.Finally,the simulation environment modeling is established by grid method,and simulation experiments are carried out on multiple maps with different complexity levels.The comparison experiments verify that traditional Q-learning is difficult to perform good path planning in large state space,and deep Q-learming enables good path planning in complex state environments.