Far and proximity maneuvers of a constellation of service satellites and autonomous pose estimation of customer satellite using machine vision

Abstract

Space robotics has a substantial interest in achieving on-orbit satellite servicing operations autonomously, e.g. rendezvous and docking/berthing (RVD) with customer and malfunctioning satellites. An on-orbit servicing vehicle requires the ability to estimate the position and attitude in situations whenever the targets are uncooperative. Such situation comes up when the target is damaged. In this context, this work presents a robust autonomous pose system applied to RVD missions. Our approach is based on computer vision, using a single camera and some previous knowledge of the target, i.e. the customer spacecraft. A rendezvous analysis mission tool for autonomous service satellite has been developed and presented, for far maneuvers, e.g. distance above 1 km from the target, and close maneuvers. The far operations consist of orbit transfer using the Lambert formulation. The close operations include the inspection phase (during which the pose estimation is computed) and the final approach phase. Our approach is based on the Lambert problem for far maneuvers and the Hill equations are used to simulate and analyze the approaching and final trajectory between target and chase during the last phase of the rendezvous operation. A method for optimally estimating the relative orientation and position between camera system and target is presented in detail. The target is modelled as an assembly of points. The pose of the target is represented by dual quaternion in order to develop a simple quadratic error function in such a way that the pose estimation task becomes a least square minimization problem. The problem of pose is solved and some methods of non-linear square optimization (Newton, Newton�Gauss, and Levenberg�Marquard) are compared and discussed in terms of accuracy and computational cost.