Data-Driven Event-Triggered Predictive Post-Impact Control of Space Robot with Uncertainties

Satellite-mounted robotic systems play an important role in various on-orbit services. The present work focuses on controlling a multi-arm space robot in the post-impact phase under the presence of uncertainties, which may arise due to dynamical parameter variation, joint friction, and disturbances. This work proposes a Gaussian Process-based Model Predictive Control (GP-MPC) to achieve the desired motion by mitigating the effects of uncertainties. Whereas the GP helps in predicting uncertainties from collected data, the MPC provides control inputs subject to the cost function and constraints. Further, an event-triggered control strategy is employed to reduce the computational burden associated with GP-MPC. It uses a Chernoff-bound-based triggering threshold to decide the occurrence of events. The generated control inputs may disturb the base’s attitude, and reactionless manipulation is also presented as a special case to mitigate it. The analytical outcomes are verified through numerical simulations on a dual-arm planer space robot.