摘要
针对农业巡检机器人在复杂作业环境中路径规划效率低、路径较长以及动态避障能力不足的问题,提出一种基于改进的深度确定性策略梯度算法(EDDPG)的路径规划方法。该方法引入旋量法(Screw Theory)优化路径规划效率,并结合人工势场法与A^(*)算法构建奖励函数,以提升动态避障能力并缩短最优路径长度。试验结果表明,在二维环境下,与对比算法相比,改进DRL算法的路径长度最短、路径平滑度最佳。在障碍物率为20%、40%、60%的条件下,其路径规划成功率分别达到0.936、0.931、0.918,均优于对比方法。此外,旋量法的使用频率越高,路径规划成功率越高。相比于其他路径规划模型,改进DRL算法的最短路径长度最大减少27.23%,平均规划时间最多缩短24.16 s。
To address the issues of low path planning efficiency,long path length,and insufficient dynamic obstacle avoidance capability in agricultural inspection robots operating in complex environments,an improved Deep Deterministic Policy Gradient(EDDPG)-based path planning method is proposed.This method incorporates Screw Theory to enhance path planning eficiency and integrates the Artificial Potential Field method with the A*algorithm to construct a reward function,thereby improving dynamic obstacle avoidance and reducing the optimal path length.Experimental results demonstrate that,in a two-dimensional environment,the improved DRL algorithm achieves the shortest path length and the best path smoothness compared to baseline algorithms.Under obstacle rates of 20%,40%and 60%,its path planning success rates reach o.936,o.93l and o.918,respectively,all outperforming the compared methods.Furthermore,a higher frequency of Screw Theory application leads to an increased path planning success rate.Compared to other path planning models,the improved DRL algorithm achieves a maximum reduction of 27.23%in the shortest path length and decreases the average planning time by up to 24.16 s.
作者
许寒琪
寇志伟
李欣
李鄂阳
Xu Hanqi;Kou Zhiwei;Li Xin;Li Eyang(College of Electric Power,Inner Mongolia University of Technology,Hohhot,010051,China;Inner Mongolia Key Laboratory of Electromechanical Control,Hohhot,010080,China)
出处
《中国农机化学报》
北大核心
2026年第4期187-194,F0003,共9页
Journal of Chinese Agricultural Mechanization
基金
国家级大学生创新创业训练计划项目(202410128013)
内蒙古科技计划项目(2021GG0256)。
关键词
农业自动化
巡检机器人
路径规划
深度强化学习
智能算法
agricultural automation
inspection robot
path planning
deep reinforcement learning
intelligent algorithm
