SGP-Based Stochastic Predictive Control of Free-Floating Space Robot in Precapture Phase

This work focuses on controlling satellite-mounted robots during the precapture phase, subject to uncertainties in dynamical parameters, joint friction, and external disturbances. We propose a sparse Gaussian process-based stochastic model predictive control (SGP-SMPC) scheme to address these challenges. The SGP-SMPC leverages the SGP technique to estimate uncertainties using previous operational data with low computational costs. This approach provides a more accurate dynamic model to effectively manage uncertainties, thereby enhancing the capabilities of SMPC and ensuring robust control performance. Numerical simulations using a planar free-floating space robot validate and demonstrate the effectiveness of the proposed method. In addition, the reactionless manipulation constraint is integrated into the SGP-SMPC framework to reduce attitude disturbances in the satellite caused by the motion of the robot arm.