Synergistic learning and observation for robust attitude control of flexible spacecraft

Abstract

Tomographic synthetic aperture radar (TomoSAR) missions require precise attitude control in the presence of low-frequency, weakly damped vibrations from large flexible appendages. To overcome the conservatism of model-based controllers and the lack of safety guarantees in data-driven approaches, this study aims to propose a control framework that integrates prescribed-time control with a learning-observation collaborative mechanism. The architecture uses a prescribed-time time-varying sliding-mode controller to ensure tracking error convergence within a user-defined interval. A Gaussian process (GP) regression model is used for online learning and feedforward compensation of structured uncertainties. A fixed-time disturbance observer runs continuously, with its compensation selectively engaged when the GP residual exceeds a threshold, enabling rapid compensation for residuals and abrupt disturbances. Lyapunov analysis establishes the prescribed-time stability of the closed-loop system. Simulations show that the proposed method reduces the attitude error to approximately 10−5 rad and achieves effective suppression of flexible vibrations compared with standard sliding mode controllers. This study offers a collaborative framework that addresses the limitations of both conservative model-based control and unverified data-driven approaches, providing an effective solution for high-precision three-dimensional TomoSAR imaging in space environments.