ROBOTIC CAPTURE AND DE-ORBIT OF A HEAVY , UNCOOPERATIVE AND TUMBELING TARGET SATELLITE IN LOW EARTH ORBIT

This paper presents a robotic capture concept for large spacecraft in low Earth orbit (LEO) that was developed as part of the the e.Deorbit feasibility study within the scope of the clean space initiative of the European Space Agency (ESA). The defective and tumbling satellite ENVISAT has been chosen as potential target to be captured, stabilized, and subsequently de-orbited in a controlled manner. Following a thorough target analysis including potential grasping points, a robotic capture concept was developed that is based on a 7-DoF dexterous robotic manipulator, a linear two-bracket gripper, and a clamping mechanism for achieving stiff fixation between target and chaser satellites prior to the de-tumbling and execution of the de-orbit maneuver. The robotic grasp concept includes a stereo-vision camera system featuring a visual servoing algorithm for camera-in-the-loop error correction. In addition, a platform-mounted camera system is utilized for target model building as well as relative motion and pose estimation. For concept validation, visual servoing, haptic grasp and capture simulations were performed. The task-specific kinematics of the manipulator and potential joint locks as contingency events were validated and analyzed using the method of capability maps. For the complete robotic capture maneuver, an error budget was created and evaluated. Geometric analysis and haptic grasp simulations showed that the gripper design and connected grasp approach is feasible and robust. In addition, the kinematics analysis yielded a sufficient reachability, even in the case of an improbable joint lock. FE-analyses were performed to show that a high compound stiffness can be achieved by the clamping device and that all required forces can be transmitted without damaging the target’s structure. Overall, the study showed that the capture and deorbiting of ENVISAT is feasible and robust using the developed robotic concept.