摘要
针对传统Q-learning算法出现的规划路线转折点多,探索效率低,以及无法实现动态环境下的路径规划问题,提出一种基于改进Q-learning算法和动态窗口法(DWA)的融合算法。首先,改变传统Q-learning算法的搜索方式,由原先的8方向变成16方向;利用模拟退火算法对Q-learning进行迭代优化;通过路径节点优化算法进行节点简化,提高路径平滑度。然后,提取改进Q-learning算法规划路径的节点,将其作为DWA算法的临时目标,前进过程中,能够实时躲避环境中出现的动静态障碍物。最终实验结果表明:融合算法具有较好的路径规划能力,实现了全局最优和有效避障的效果。
Aiming at the problems of traditional Q-learning algorithm with many turning points in the planning route,low exploration efficiency,and the inability to achieve path planning in dynamic environment,a fusion algorithm based on improved Q-learning algorithm and dynamic window approach(DWA)is proposed.The search method of traditional Q-learning algorithm is changed from the original 8 directions to 16 directions.The simulated annealing algorithm is used to iteratively optimize Q-learning.The path node optimization algorithm is used to simplify the nodes and improve the smoothness of the path.Then the nodes of the improved Q-learning algorithm planning path are extracted as the temporary target points of DWA algorithm.In the process of moving forward,the dynamic and static obstacles in the environment can be avoided in real time.The experimental results show that the fusion algorithm has better path planning ability,and achieves the effect of global optimum and effective obstacle avoidance.
作者
王志伟
邹艳丽
刘唐慧美
侯凤萍
余自淳
WANG Zhiwei;ZOU Yani;LIU Tanghuimei;HOU Fengping;YU Zichun(College of Electronic Engineering,Guangxi Normal University,Guilin 541004,China)
出处
《传感器与微系统》
CSCD
北大核心
2023年第9期148-152,共5页
Transducer and Microsystem Technologies
基金
广西重大科技专项项目(桂科AA21077015)
广西多源信息挖掘与安全重点实验室系统性研究课题基金资助项目(13—A—02—03)。