Research/ IN-SPACE SERVICING
Towards Safe On Board Autonomy in Robotic On Orbit Servicing
Peter Kötting, Alexander Meinert, Peter Stadler, Niklas Baldauf, Fabio Schneider, Roberto Lampariello
Guidance, Dynamics & ControlOpenAlex topic: Advanced Control Systems Optimization · Motion Planning & NavigationOpenAlex keyword: collision avoidance · Autonomy, Task Planning & OperationsOpenAlex keyword: onboard autonomy
Safe on-board motion control for robotic On-Orbit Servicing is challenging due to the cou-pled multibody dynamics of the servicer and its operational safety constraints. This paperpresents and evaluates control approaches based on Model Predictive Control and Con-strained Imitation Learning for on-board motion control. MPC formulations are developedfor different control architectures of increasing complexity, ranging from relative-positionand spacecraft-attitude control to a combined control scheme. We address stability, recursivefeasibility, actuator limits, joint constraints, collision avoidance, and field-of-view require-ments. The constrained imitation-learning approach approximates the optimal controlpolicy with a neural network and incorporates control-barrier- and control-Lyapunov-function-based safety and stability certificates during training and deployment to overcomethe computational burden of solving the MPC online. As a benchmark case, we consider adetumbling scenario of ESA’s Envisat satellite. The control frameworks are evaluated in ahigh-fidelity Modelica-based simulation environment following a representative concept ofoperations. Performance metrics include propellant consumption, computational runtime,and constraint violations.