Deep reinforcement learning-based trajectory planning with continuous pose representation for 6-DoF free-floating space robot

Trajectory planning for free-floating space manipulator is a challenging task due to the complex coupling between the manipulator and its floating base. In this paper, we propose the RTPC algorithm, a reinforcement learning-based trajectory planning method to address the 6-DoF free-floating space manipulation tasks. To overcome the discontinuity and non-uniqueness of conventional pose representations, we introduce a continuous formulation based on transformation matrix and incorporate it into the state space, enabling more stable learning and improving trajectory planning performance. To address the challenges of efficient convergence and enhance exploration in the early and mid-stages of training, we designed the Hybrid Efficient Learning (HybridEL) framework, which combines hindsight experience replay with precision-based curriculum learning. We conduct comprehensive experiments to validate the effectiveness of our proposed algorithm. The results demonstrate that our approach achieves better convergence and superior performance in space manipulator trajectory planning tasks compared to existing methods, while maintaining real-time capability. The code and simulation configuration for this work are publicly available at https://github.com/HIT-YuhuiHu/RTPC.