基于强化学习的自动驾驶汽车路径规划方法研究综述  

Review of research on path planning methods for autonomousvehicles based on reinforcement learning

作　　者：许宏鑫 吴志周 梁韵逸[3] Xu Hongxin;Wu Zhizhou;Liang Yunyi(College of Mechanical Engineering,Xinjiang University,Urumqi 830017,China;College of Transportation Engineering,Tongji University,Shanghai 201804,China;School of Engineering&Design,Technical University of Munich,Munich 999035,Germany)

机构地区：[1]新疆大学机械工程学院,乌鲁木齐830017 [2]同济大学交通运输工程学院,上海201804 [3]慕尼黑工业大学工程与设计学院,德国慕尼黑999035

出　　处：《计算机应用研究》2023年第11期3211-3217,共7页Application Research of Computers

基　　金：国家自然科学基金资助项目(52172330,52002281)

摘　　要：路径规划作为自动驾驶的关键技术,具有广阔的应用前景和科研价值。探索解决自动驾驶车辆路径规划问题的方法,着重关注基于强化学习的路径规划方法。在阐述基于常规方法和强化学习方法的路径规划技术的基础上,重点总结了基于强化学习和深度强化学习来解决自动驾驶车辆路径规划问题的算法,并将算法按照基于值和基于策略的方式进行分类,分析各类算法的特点、优缺点及改进措施。最后对基于强化学习的路径规划技术的未来发展方向进行了展望。As a key technology of autonomous driving,path planning has broad application prospects and scientific research value.This paper explored ways to solve the path planning problem for autonomous vehicles,focusing on reinforcement lear-ning-based path planning methods.On the basis of expounding the path planning technology based on conventional method and reinforcement learning method,this paper focused on summarizing the algorithm based on reinforcement learning and deep reinforcement learning method to solve the path planning problem of autonomous vehicles,classified the algorithm according to value-based and policy-based methods,analyzed the characteristics,benefits and drawbacks,and improvement measures of each type of algorithms.Finally,this paper looked forward to the future development direction of path planning technology based on reinforcement learning.

关 键 词：自动驾驶 路径规划 强化学习 马尔可夫决策过程 

分 类 号：TP181[自动化与计算机技术—控制理论与控制工程]