Collision monitoring within satellite clusters

An autonomous online collision avoidance methodology for satellite clusters is developed and simulated. Contours of probability, defined using state uncertainty ellipsoids, are used to monitor and predict potential collisions over time. The methodology integrates three approaches in order to reduce online computation for autonomous operations while still producing accurate predictions: 1) monitoring distances between bounded probability ellipsoids, which are used to evaluate when and how fast a collision may occur, 2) monitoring ellipsoids as they intersect, then calculating an outer bound of collision probability, and 3) numerical evaluation of collision probability using a three-dimensional (3-D) convolution integral to evaluate conservatism and insure accuracy. The approach assumes the initial state error (relative position and velocity), sensor error, and disturbance function are bounded by ellipsoids to a known level of probability. The methodology can be applied in the presence of different control schemes and system faults. If a collision is imminent, a control reconfiguration can then occur in order to move the healthy satellites to parking orbits. Simulation results show the proposed methodology to not be overly conservative. In addition, results show that satellite clusters with a close proximity will require a backup system that is accurate and gives measurement updates at least every few minutes.