3D Path Planning Digital Twins for Robot Arms Incorporating Reinforcement Learning

Aiming at the problems of low motion efficiency of the robotic arm in the space full of obstacles and the inability to monitor the operation status in real time, this paper firstly designs the digital twin system architecture to realize the virtual reality mapping and interaction. Secondly, considering the requirement of robotic arm path planning, a digital twin-based APF-SAC path planning method is proposed, which optimizes the path generated by the artificial potential field method as the reference path and incorporates it into the reward function to accelerate the convergence speed of the traditional SAC algorithm, and introduces the recent experience playback technology to improve the stability and speed of training. Experiments based on the proposed algorithm show that the path imitation mechanism can improve the training efficiency, and the APF-SAC algorithm can plan the path more reasonably compared with the PPO algorithm. On this basis, the digital twin system is used to realize the real-time preview of the algorithm, which ensures the availability of the path and the safety of the robotic arm.