Rendezvous with a Non-Cooperating Target

Abstract

Space robotics has a substantial interest in achieving on-orbit service (OOS) operations autonomously, i.e rendezvous and docking/berthing (RVDB) with failed, stranded or malfunction satellites. Visual navigation system has a wide area of applications. In the context of this thesis we investigate a visual navigation system for on-orbit service (OOS). The space servicing area has a very important appeal in space flight dynamics mainly in these days when the scientific community is deeply concerned with space debris and the risk those objects impose on other space missions and even on the ecosystem not to say risk on human life on ground. Moreover current database indicates regular opportunities for satellite servicing including on-orbit upgrades, repair and rescue of stranded satellites. Those satellites are classified as non-cooperative targets when under the goal of space cleaning or space maintenance by space tugs. A chaser spacecraft would have to have the capability to rendezvous and grasp a dead satellite or other space debris, taking it out of its orbit for the sake of fixing problems, refueling, or just for cleaning purposes, without any cooperation from the target vehicle. In this sense failed or dead satellites become non-cooperative targets that are not able to provide any information on their position and attitude and eventually are not capable of maneuvering to cooperate with the docking operation. Position and attitude here shall be understood as relative orbital position and relative attitude between the chaser and target space vehicles or chaser vehicles and other space objects. This thesis presents an algorithm developed for estimating the pose (position and attitude) and the motion (velocities) of a generic target satellite based on visual navigation. The algorithm gathers the strength of classical attitude estimation methods to obtain real-time applicability conditions when adopting a Kalman filter for sequential state estimation. The visual system is monocular and satellite model-based. Therefore, it does not rely on any marker attached to the target satellite. The navigation solution can be used for a vast category of applications such as space debris removal, servicing for stranded satellites, and interception of hostile objects. The approach is first tested with synthetic image data from a spacecraft object generated in virtual reality. A test-bed is used to simulate the on-orbit optical environment. A scaled satellite model and a CCD camera is employed in the test bed in order to evaluate the algorithm for real-time application. A combination of three dimensional (3D) model-based attitude estimation and filtering time series of images produce similar real time solution for the estimation problem and increases the reliability of the relative attitude and position results.