基于强化学习的无人车组路径优化算法研究
Research on Path Optimization Algorithm of Unmanned Vehicle Group Based on Reinforcement Learning
司炳山 1董志明 2孙茂凡3
作者信息
- 1. 陆军装甲兵学院,北京 100071;中国人民解放军 75130 部队,广西 贵港 537100
- 2. 陆军装甲兵学院,北京 100071
- 3. 中国航天科工集团第二研究院,北京 100000
- 折叠
摘要
针对传统单车路径规划算法在进行无人车组路径规划时存在的算法收敛性问题,采用强化学习方法,对传统 Q-learning算法中的探索率进行改进,将每一个路程点作为每一段局部路径规划的目标点,通过传感器感知外界环境的信息,进行基于强化学习的在线局部路径规划,完成避障和寻径任务.构建了算法模型与仿真环境,并进行了仿真,结果表明无人车组能够在短时间内收敛到稳定状态并自主完成规划任务,证明了算法的有效性和可行性.上述算法在多无人战车协同的智能规划与控制中具有良好的应用前景.
Abstract
In response to the convergence problem of traditional bicycle path planning algorithms in unmanned ve-hicle group path planning,this paper adopts reinforcement learning method to improve the exploration rate in the tradi-tional Q-learning algorithm.Each distance point was used as the target point for each local path planning segment,and sensors were used to perceive information from the external environment for online local path planning based on reinforcement learning,completing obstacle avoidance and path finding tasks.The simulation environment of the algo-rithm and the simulation experiment were constructed,and the results show that the unmanned vehicle group can con-verge to the stable state in a short time and complete the planning task autonomously,and prove the validity and feasi-bility of the algorithm.The algorithm has good application prospect in intelligent planning and control of multi-un-manned vehicles.
关键词
强化学习/无人战车/路径优化/探索率Key words
Reinforcement learning/Unmanned vehicle/Path optimization/Exploration rate引用本文复制引用
出版年
2024
NETL
NSTL
万方数据