Intelligent Control Method of Robotic Arm Follow-Up Based on Reinforcement Learning

The preparation of lunar soil samples is a complex task, and direct programming control of the robotic arm is too cumbersome and not safe enough under the existing conditions. Therefore, this paper proposes a six-degree-of-freedom robotic arm follow-up operation method based on human motion characteristics and deep reinforcement learning to improve the efficiency and safety of the preparation process. Firstly, the motion data of human arm under different tasks collected by Optitrack motion capture system are analyzed to get the motion law of joint angle, and the corresponding reward function is designed. Then, the robotic arm is trained in the simulation environment using Soft Actor-Critic (SAC) and Hindsight Experience Replay (HER) algorithms. According to the trained network model, the robot arm completes the task of tracking random target motion. Finally, the experiments in Pybullet simulation environment verify the feasibility, effectiveness and adaptability of the manipulator control method proposed in this paper. The method can be applied to other continuous motion control scenarios in continuous space.