Globally Convergent Near-Optimal Pose and Inertia Estimation of On-Orbit Uncooperative Targets

Rendezvous with uncooperative targets, such as space debris, inoperative satellites, and asteroids requires flight synchronization between the chaser spacecraft and the target. Therefore, the relative attitude and position, as well as the relative angular and linear velocities need to be estimated using relative navigation sensors, such as cameras and LIDAR. This work presents a filtering solution for on-orbit relative pose estimation of uncooperative targets that, resorting to measurements of the relative position and velocity of features on the target and given rich target rotational trajectories, offers convergence guarantees, near-optimal performance, and estimation of the target's inertia tensor up to a scale factor. The solution is inspired in a recent filtering methodology of dynamic systems called the eXogenous Kalman filter (XKF), which is a cascade of a globally convergent observer with a filter similar to the extended Kalman filter (EKF). In the latter, the difference lies in the linearization of the nominal system about the globally convergent estimate. The proposed solution inherits the advantages of both filters without inheriting their individual drawbacks. Finally, the performance and stability are thoroughly evaluated using Monte Carlo simulations. An EKF is implemented for the same problem for comparison purposes. The results show that the performance of the XKF is similar to the performance of the EKF. However, the XKF offers global convergence guarantees, crucial to increase trust in the algorithm, something the EKF does not.