A Learning-Based Scheme for Safe Deployment of Tethered Space Robot

This work focuses on the problem of collision avoidance with space debris for large-scale deployment of a tethered space robot (TSR). To this end, a general scheme that contains offline training and online execution is presented for safe deployment of a TSR. Specifically, inspired by the contraction theory, a feedback controller is learned from data to guarantee the superior tracking performance in the offline phase. Furthermore, the “tube” where state of the TSR would stay within is optimized simultaneously. In the online execution phase, when the space debris are detected, the motion planner generates a nominal trajectory by considering safety constraints. Then, in the presence of disturbances, the feedback controller learned offline tracks this nominal trajectory safely without collisions. The proposed scheme allows for the comprehensive utilization of prior knowledge for designing the tracking controller in the offline phase, thereby enhancing the online tracking performance. Finally, the numerical simulations demonstrate effectiveness of the proposed framework.