一种基于DDPG算法的6轴机械臂控制研究

Research on the control of a six-axis robotic arm using the DDPG algorithm

针对传统控制算法在复杂环境下,精度低、稳定性不足等问题,提出了一种基于深度确定策略梯度算法(deep deterministic policy gradient,DDPG)的控制算法,以更好地解决6轴机械臂在三维空间中的控制难题。在MuJoCo平台上建立仿真环境,引入所设计机械臂为测试对象,并采用DDPG算法、柔性动作-评估算法(soft actor-critic algorithms,SAC)和双延迟深度确定策略梯度算法(twin delayed deep deterministic policy gradient,TD3)对仿真环境下的机械臂进行了几组对比测试。研究表明:以DDPG算法为基础的机械臂控制方法能够有效地提高机械臂的控制精度,相对于SAC、TD3等算法稳定性较好。

A reinforcement learning method based on Deep Deterministic Policy Gradient(DDPG)is presented to more effectively tackle the problem of controlling a six-axis robotic arm in three-dimensional space in order to address the issues of low accuracy,stability,and executability of existing control algorithms in complicated situations.The simulation environment is established in the MuJoCo platform,the planned robotic arm is imported as the test object,and the DDPG algorithm,the Soft Actor-Critic Algorithms(SAC),and the Twin Delayed Deep Deterministic Policy Gradient(TD3)are utilized for repeated comparison tests in the simulation environment.The study demonstrates that the DDPG algorithm-based robotic arm control approach can successfully increase the accuracy and stability of robotic arm control,and that this algorithm is more stable than SAC and TD3.