Attitude Maneuver Control of Rigid Spacecraft Based on Model-Free Controller Guide SAC Algorithm

This paper presents a model-free attitude control algorithm applied to rigid spacecraft. The Soft Actor–Critical (SAC) deep reinforcement learning algorithm is initially employed for spacecraft control, and two modifications are proposed to address the issues of slow training speed and poor convergence accuracy. The modifications encompass the PID-Guided SAC algorithm, which integrates a PID controller with SAC to optimize control strategies and enhance performance, and the adaptive TD3-SAC algorithm, which dynamically adjusts the algorithm weights using a [Formula: see text]-greedy strategy to develop a comprehensive hybrid control strategy. This process increases the stability, and convergence accuracy of both algorithms. Ultimately, the experimental results validate the effectiveness and superiority of these algorithms in spacecraft attitude maneuver control.