基于深度强化学习的移动机器人路径规划研究
Research on Path Planning of Mobile Robots Based on Deep Reinforcement Learning
荣垂霆 1朱恒伟 1张宾 1刘聪1
作者信息
摘要
鉴于采用深度强化学习算法进行移动机器人路径规划时存在收敛速度慢的问题,提出一种改进的算法.对经验回放机制中样本的学习潜力得分进行设计,根据学习潜力得分对样本进行优先级评分,并根据评分进行采样.将改进算法应用到机器人路径规划任务中,并进行奖励函数、避障参数及路径规划实验环境的设计.通过与对比算法进行实验比较,验证了改进算法的收敛速度及其在路径规划任务中的有效性.
Abstract
Given the problem of slow convergence speed when using Deep Reinforcement Learning algorithms for mobile robot path planning,an improved algorithm is proposed.It designs the learning potential score of samples in the experience replay mechanism,prioritizes the samples based on the learning potential score,and samples them according to the score.It applies improved algorithms to robot path planning tasks and designs reward functions,obstacle avoidance parameters,and path planning experimental environments.Through experimental comparison with comparative algorithms,the convergence speed of the improved algorithm and its effectiveness in path planning tasks are verified.
关键词
深度强化学习/路径规划/移动机器人Key words
Deep Reinforcement Learning/path planning/mobile robot引用本文复制引用
出版年
2024
国家科技期刊平台
NSTL
万方数据