Analysis and optimization of PMD-sensor data for rendezvous applications in space

Kurzfassung

In the last decade, people became more aware of the critical situation in space concerning the space debris. Nowadays, there are plenty of uncontrolled objects located on different orbits in the outer space. Some of them present the real hazards to the functioning satellites and to the International Space Station. The researches all over the world are working on optimal solutions for the space cleaning. There are projects which are directed to repair the satellites and extend their operational lifetime, or totally remove a no more useful space object from its orbit. Rendezvous is the one inevitable space operation all these tasks need. We are able to execute a space operation with a target satellite only if the servicer satellite approaches it. The goal of this thesis is to develop a visual navigation system with the Photonic Mixer Device (PMD) sensor for the close rendezvous phase with a non-cooperative target. The PMD sensor has never been used in space missions so far, but it has already been tested on the ground. In order to achieve the goal, this thesis provides two different pipelines for the pose acquisition and for the pose tracking using the images from the PMD sensor. In this work we show that the use of the PMD's amplitude and depth images together brings a great contribution to the visual navigation system. The pose acquisition is required in order to initialize a pose (position and orientation) of the non-cooperative space object before the tracking takes place. This task is very diffcult, because we have to deal only with the PMD images and the known 3D model of the target. We initialize the pose with the depth image and thereafter improve the obtained pose with the amplitude image. As soon as we have an initial estimation, the servicer starts to approach the target in a frame-to-frame mode. The pose for every frame is estimated by a fusion of the states, which are calculated with the developed techniques for the depth and the amplitude images correspondingly. This technique shows a stable tracking with low errors of the estimated pose even if there are some distortions in the depth image. This fact is very important for the close rendezvous phase, because any failures in Guidance, Navigation and Control (GNC) system can lead to an unpredictable behavior of the chaser, and in end effect, create more space debris. The techniques presented in this thesis are tested with real images of the PMD sensor. The rendezvous simulations are executed and evaluated with a high accuracy hardware-inthe- loop simulator. The tests of the visual navigation with a PMD sensor in a closed loop show promising results. The servicer satellite can smoothly approach the non-cooperative target by only using the PMD sensor for relative navigation.