Adaptive Relative Spacecraft Navigation in Proximity Operations with Unmodeled Target Behavior

Abstract

Abstract An online method is presented for detecting and estimating unmodeled translational maneuvers of a target vehicle. Designed for relative navigation scenarios, the algorithm runs onboard the chaser vehicle and operates without communication from the target. By leveraging a buffer of observed pre-fit measurement innovations, the method eliminates the need for prior knowledge about the timing, magnitude, or direction of the target’s maneuvers. This method integrates adaptive process noise estimation with a splitting algorithm designed specifically for this type of unmodeled dynamics. The algorithm demonstrates rapid detection and estimation of unmodeled target maneuvers in a low Earth orbit scenario, without relying on inflated uncertainty bounds. Its performance surpasses that of state-of-the-art adaptive filters designed with prior knowledge of the maneuver. The algorithm also demonstrates competitive performance in detecting and estimating target maneuvers in a cislunar scenario, matching or exceeding state-of-the-art filters configured with prior knowledge of the maneuver. In both scenarios, the adaptive estimation of process noise is key to enable both rapid transient response to detected maneuvers and accurate steady-state performance; all without any prior knowledge of the target’s maneuver timing, magnitude and direction.