衰减高斯噪声DDPG算法的机械臂轨迹规划  

Trajectory planning of robotic arm based on the gaussian decaying noise DDPG algorithm

作　　者：周雨溪 赵慧[1,2] 韩晓峰 ZHOU Yuxi;ZHAO Hui;HAN Xiaofeng(Hubei Province Key Laboratory of Mechanical Transmission and Manufacturing Engineering,Wuhan University of Science and Technology,Wuhan 430081,Hubei,China;Institute of Robotics and Intelligent Systems,Wuhan University of Science and Technology,Wuhan 430081,Hubei,China)

机构地区：[1]武汉科技大学湖北省机械传动与制造工程重点实验室,湖北武汉430081 [2]武汉科技大学机器人与智能系统研究院,湖北武汉430081

出　　处：《农业装备与车辆工程》2024年第10期111-118,共8页Agricultural Equipment & Vehicle Engineering

摘　　要：针对农业采摘机械臂的DDPG算法轨迹规任务中,调查了因高斯噪声标准差取值不当导致的强化学习训练失败问题,提出一种衰减正态噪声的DDPG算法,使高斯标准差σ随训练回合数增加而减小;利用Mujoco物理引擎进行多次仿真训练,验证衰减正态噪声相较于传统正态噪声在轨迹规划任务中的优势。结果表明,改进后的算法在完成采摘机械臂的轨迹规划任务时更为有效,成功解决了存在的问题。The issue of DDPG algorithm training failure caused by inappropriate values of Gaussian noise standard deviation in the trajectory planning task of agricultural picking robotic arms was investigated.To address this problem,a decaying normal noise DDPG algorithm was proposed,where the Gaussian standard deviationσdecreases as the number of training episodes increases.Multiple simulation training sessions using the Mujoco physics engine were conducted to verify the advantages of decaying normal noise over traditional normal noise in trajectory planning tasks.The results showed that the improved algorithm was more effective in completing the trajectory planning task of the picking robotic arm,successfully solving the problem.

关 键 词：强化学习 DDPG算法 正态噪声 机械臂 轨迹规划 

分 类 号：TP241[自动化与计算机技术—检测技术与自动化装置]