Pose Determination for Malfunctioned Satellites Based on Depth Information

Autonomous on-orbit servicing is the future space activity which can be utilized to extend the satellite life. Relative pose estimation for a malfunctioned satellite is one of the key technologies to achieve robotic on-orbit servicing. In this paper, a relative pose determination method by using point cloud is presented for the final phase of the rendezvous and docking of malfunctioned satellites. The method consists of three parts: (1) planes are extracted from point cloud by utilizing the random sample consensus algorithm. (2) The eigenvector matrix and the diagonal eigenvalue matrix are calculated by decomposing the point cloud distribution matrix of the extracted plane. The eigenvalues are utilized to recognize rectangular planes, and the eigenvector matrix is the attitude rotation matrix from the sensor to the plane. The solution of multisolution problem is also presented. (3) An extended Kalman filter is designed to estimate the translational states, the rotational states, the location of mass center, and the moment-of-inertia ratios. Because the method only utilizes the local features without observing the whole satellite, it is suitable for the final phase of rendezvous and docking. The algorithm is validated by a series of mathematical simulations.