This paper presents a pose estimation procedure for tracking attitude and position of a noncooperative tumbling spacecraft during rendezvous and docking maneuvers. The key aspect of the method is that the current state of the target spacecraft is estimated in real-time using a low-resolution depth-image of the scene and a known model of the vehicle. The proposed procedure exploits the iterative closest point algorithm implemented in closed-loop fashion. In order to guarantee stability of the solution, the current pose of the target is predicted onboard by propagating the last estimated state using a dedicated dynamics model of the maneuver. The tracking procedure is initialized with a fast template matching algorithm that determines attitude and position of the target without prior information on its state. The capabilities of the proposed procedure are demonstrated using a full degrees-of-freedom hardware simulator. In these experiments, the Microsoft Kinect v2 sensor is employed for real-time acquisition of the target spacecraft depth image. The results demonstrate the effectiveness of the proposed procedure during docking maneuvers to a noncooperative tumbling target However, prediction of the current target pose is crucial for algorithm robustness.
展开▼
