摘要
针对蛇形机械臂控制问题,提出了一种基于深度强化学习的控制策略,该控制策略采用深度确定性策略梯度算法(DDPG)。分析了蛇形机械臂的结构和工作范围。基于Python语言,使用gym中的pyglet模块搭建用于产生数据的仿真环境,设置奖励函数、状态变量和动作变量,最终实现了对蛇形机械臂的精确控制。仿真实验表明:DDPG算法在蛇形机械臂的控制过程中能快速收敛,同时该控制策略在2D平面可实现对目标物的快速精确逼近,并具有较好的鲁棒性。
Aiming at the control problem of snake-like arm,a control strategy based on deep reinforcement learning is proposed,which adopts deep deterministic policy gradient(DDPG).This paper analyzes the structure and working range of the snake-like arm.Based on Python language,using pyglet module in gym to build a simulation environment for generating data,setting reward function,state variables and action variables,the precise control of the snake-like arm is finally realized.The simulation results show that the DDPG algorithm can converge quickly in the control process of the snake-like arm,and the control strategy can achieve fast and accurate approximation of the target object in the 2D plane,and has good robustness.
作者
唐超
张帆
王文龙
李徐
Tang Chao;Zhang Fan;Wang Wenlong;Li Xu(School of Mechanical and Automotive Engineering,Shanghai University of Engineering Science,Shanghai 201620,China)
出处
《农业装备与车辆工程》
2022年第8期17-21,共5页
Agricultural Equipment & Vehicle Engineering
基金
上海市科委生物医药领域科技支撑计划资助(17441901200)。