Physics-Informed Gaussian Process-Based Attitude Stabilization Control for Liquid-Filled Spacecraft

This paper investigates the attitude stabilization problem for liquid-filled spacecraft. The problem is inherently challenging due to significant liquid sloshing and fuel consumption, which introduce nonlinear uncertainties and a time-varying center of mass. To address this issue, first, we establish the attitude dynamics for liquid-filled spacecraft based on the moving pulsating ball model. Then, an adaptive Physics-informed Gaussian Process Regression (PI-GPR)-based attitude stabilization strategy is pro posed, which leverages the angular momentum theorem to learn model uncertainties serving as semi-feedforward compensation. By modeling the angular momentum with GP with physics prior knowledge, the proposed method enhances the model interpretability and augments small-sample modeling efficacy. Furthermore, an adaptive updating law is introduced to estimate the shifting center of mass resulting from liquid slosh and fuel consumption. The rigorous theoretical proofs for the probabilistic Lyapunov stability and boundedness of the proposed closed-loop system are provided. Simulation study demonstrates the effectiveness and high performance of the proposed strategy.