Relative navigation of uncooperative space bodies

This paper develops a novel approach for relative navigation of two space bodies for cases in which communication with the target space body is either inconvenient or impossible. The approach utilizes multiple hypothesis tracking (MHT) to track the features detected from optical image data using standard edge detection techniques. This is in contrast to traditional optical-only relative navigation systems that use other approaches such as machine learning techniques or optical flow techniques to track features. In machine learning, the algorithms must first be trained through an extensive set of testing scenarios to learn an end-to-end solution, and must be retrained for every particular target model-frame. Optical flow algorithms also have their own inherent limitations, such as dealing with discontinuities of motion across object boundaries in the scene. MHT uses state estimate information within its decision logic, allowing it to make data association decisions that are more robust to relative navigation scenarios than the traditional methods. The tracked features from MHT are utilized as measurements in a multiplicative extended Kalman Filter (MEKF). Using the MEKF to feed predicted states into the MHT algorithm, which in turn provides measurements back to the MEKF, estimates of the attitude, angular rate, position and velocity of the target body are obtained.