基于强化学习的无人驾驶仿真研究  

Research on Unmanned Driving Simulation Based on Reinforcement Learning

作　　者：孙嘉浩 陈劲杰[1] Sun Jiahao;Chen Jinjie(University of Shanghai for Science and Technology,Shanghai 200093,China)

机构地区：[1]上海理工大学机械电子学院,上海市200093

出　　处：《农业装备与车辆工程》2020年第6期102-106,共5页Agricultural Equipment & Vehicle Engineering

摘　　要：提出一种基于强化学习的无人驾驶仿真方案,采用Deep Q-Learning算法,设置经验池来对驾驶策略进行学习,设计了控制策略和动作策略来控制虚拟环境下的驾驶仿真。在无人驾驶仿真平台TORCS上进行了仿真实验,对无人驾驶进行训练,训练结果验证了该算法的有效性与可行性。该强化学习算法对无人驾驶仿真提供了可行方案的参考结论。An unmanned driving simulation scheme based on reinforcement learning is proposed.Deep Q-Learning algorithm is used to set up experience pool to learn driving strategy.Control strategy and action strategy are designed to control driving simulation in virtual environment.The simulation experiment is carried out on the unmanned driving simulation platform TORCS,and the unmanned driving is trained.The training results verify the validity and feasibility of the algorithm.The conclusion that the reinforcement learning algorithm provides a feasible scheme for the unmanned driving simulation is drawn.

关 键 词：无人驾驶 强化学习 Deep Q-Learning 驾驶仿真 TORCS 

分 类 号：TP399[自动化与计算机技术—计算机应用技术]