Enhancing 3D Trajectory Tracking of Robotic Manipulator Using Sequential Deep Reinforcement Learning with Disturbance Rejection

摘要

This paper addresses the problem of trajectory tracking for robotic manipulators in three-dimensional space. We use Deep Deterministic Policy Gradient (DDPG), a model-free deep reinforcement learning technique, with a sequential training to significantly expedite the training process. The reward function is a key component in reinforcement learning. Our proposed design is particularly tailored to handle external disturbances. This feature ensures robust performance and adaptability, which is crucial for real-world applications of robotic manipulators in dynamic environments. To evaluate the effectiveness and efficiency of our approach, we present comprehensive numerical simulation results. These results not only demonstrate the capability of our model to facilitate a faster training rate but also showcase a remarkable reduction in the tracking mean square error (MSE).