Control Barrier Functions for Safe Real-Time Control of Spacecraft Onboard Robotic Manipulators

Spacecraft onboard robotic manipulators are increasingly considered for tasks ranging from active debris removal to inspection and maintenance. Real-time control of such systems remains challenging, due to the high dimensional non-linear dynamics and operational constraints. This paper presents a Control Barrier Function (CBF) based safety filter that minimally modifies arbitrary reference control signals to ensure satisfaction of workspace and collision avoidance constraints of a floating-base robot. The safety filter simultaneously ensures satisfaction of 37 CBF-induced constraints including platform angular velocity, joint limits, and selfcollision avoidance. The filter is deployed onto a simulated 16 Degree-of-Freedom (DoF) free-floating robotic manipulator, that comprises a free-floating platform, actuated in rotation via reaction wheels, and a 7-DoF robotic manipulator. Constraint satisfaction is validated by training an adversarial Reinforcement Learning (RL) controller to provide worst case reference controls. After validation the RL reference controller is trained to complete end effector full-pose Cartesian-space point-topoint manoeuvres. It is shown that the RL controller can directly map the Cartesian task space to joint space actuator commands whilst the proposed filter ensures the system’s safety.