基于神经网络Q-learning算法的智能车路径规划

Intelligent Vehicle Path Planning Based on Neural Network Q-learning Algorithm

针对智能小车行走过程中的全局路径规划和路障规避问题,提出了一种基于神经网络Q-learning强化学习算法,采用RBF(Radial Basis Function)网络对Q学习算法的动作值函数进行逼近,基于MATLAB环境开发了智能小车全局路径规划和路障规避仿真系统。与传统的以及基于势场的Q学习算法相比,所采用的算法能更加有效地完成智能小车在行驶环境中的路径规划和路障规避。仿真结果表明:算法具有更好的收敛速度,可增强智能小车的自导航能力。

A reinforcement learning algorithm based on neural network Q-learning is proposed to solve the problem of global path planning and obstacle avoidance.RBF(Radial Basis Function)network is used to approximate the action value function of Q learning algorithm.The global path planning and obstacle avoidance simulation system is developed by MATLAB.Compared with the traditional and potential field Q algorithm,the algorithm can be more effective to complete the path planning and obstacle avoidance of intelligent car in the driving environment.The simulation results show that the algorithm has better convergence speed and the ability of self navigation.