Trajectory Optimization of Rapid Space Debris Capture Based on Reinforcement Learning

With the rapid development of the space industry, space debris has become a critical challenge, posing significant threats to the safety of on-orbit spacecraft. As a result, active debris removal (ADR) has emerged as an urgent necessity. A key aspect of ADR missions is the transfer trajectory optimization, which can significantly reduce fuel consumption and improve operational efficiency. However, traditional numerical optimization methods exhibit inherent limitations in convergence and computational efficiency, particularly when handling prolonged processes governed by multifaceted constraints. To address these challenges, we proposes a reinforcement learning-based trajectory optimization method for rapid space debris capture. Our approach enables fast transfer trajectory planning based on terminal conditions while remaining robust against environmental disturbances. Experimental results indicate that, for equivalent fuel consumption and mission success rates, our algorithm achieves computation speeds up to 100 times faster than genetic algorithms (GA) and particle swarm optimization (PSO), requiring only 1% of their computation times. Furthermore, our method maintains high performance under environmental disturbances and measurement errors by continuously remeasuring and adjusting the trajectory in real time as the target is approached. This capability significantly enhances mission success rates, even as the performance of GA and PSO degrades under similar conditions.