Improving autonomous guidance using machine learning technologies

Active debris removal missions pose demanding requirements on the visual guidance system. We investigate the potential applications of machine learning technologies to solve some of the remaining challenges for these missions. A novel method of attitude determination of an unknown and uncooperative debris satellite is presented, which adopts machine learning technologies to detect and track inherently useful image landmarks. We then apply image segmentation and object detection approaches to this domain and demonstrate their advantages. The performance of the algorithms are analysed via numerical simulation and compared with conventional approaches. In order to facilitate further research into the applications of machine learning for visual guidance in space, we make available a simulation framework which is capable of generating realistic image data simulating the relative motion between co-orbiting satellites