重庆科技大学学报（自然科学版）2025，Vol.27Issue(3)：96-105,10.DOI:10.19406/j.issn.2097-4531.2025.03.010
基于深度强化学习的机械臂避障路径规划
Obstacle Avoidance Path Planning for Robotic Manipulator Based on Deep Reinforcement Learning
摘要
Abstract
To improve the obstacle avoidance path planning capability of robotic manipulator,this paper proposes a deep reinforcement learning algorithm that combines improved reward function with optimized hyperparameter.First,a reward function for the robotic manipulator is designed in obstacle environments,and tree-structured parzen estimator(TPE)algorithm is adopted for hyperparameter tuning,thereby achieving collision-free path planning for the robotic manipulator.The mathematical model of the robotic manipulator is constructed,and path planning simu-lation experiments are conducted.Simulation results show that after combining the reward function and optimized hy-perparameters,the convergence effect of the SAC algorithm is greatly improved.The effectiveness of deep reinforce-ment learning algorithms for obstacle avoidance path planning is verified in real experimental environments.关键词
机械臂/避障路径规划/深度强化学习/贝叶斯优化Key words
robot manipulator/obstacle avoidance path planning/deep reinforcement learning/bayesian optimiza-tion分类
信息技术与安全科学引用本文复制引用
朱广,陆俊谕,史艳琼,张永华,张旭..基于深度强化学习的机械臂避障路径规划[J].重庆科技大学学报（自然科学版）,2025,27(3):96-105,10.基金项目
大学生创新创业项目"智慧农业大棚综合检测管理系统"(202310878064) （202310878064）
安徽省科技重大专项"封装工艺过程3D缺陷检测仪"(202203A05020022) （202203A05020022）
无锡维度机器视觉产业技术研究院有限公司委托项目"机械手驱动三维缺陷视觉检测方法与点云数据处理算法研究"(HYB20230260) （HYB20230260）
