Title:Autonomous Angles-Only Multi-Target Tracking for Spacecraft Swarms

Abstract:Autonomous angles-only relative navigation is a key enabler towards operating spacecraft swarms in deep space. However, current architectures for angles-only relative navigation do not consider the measurement assignment problem and simply assume that measurements are correctly assigned to targets. Consequently, there is a need for new algorithms which accurately and robustly assign measurements to multiple target space objects in view, without requiring existing relative orbit knowledge. This paper presents the Spacecraft Angles-Only Multi-Target Tracking Software algorithm that performs this task using only sequential camera images and under autonomous spacecraft limitations, applying 1) kinematic knowledge of target behavior in the observer's reference frame, and 2) principles of multi-hypothesis tracking to treat ambiguous measurement assignments. A measurement transform is introduced to ensure consistent elliptical target motion, and kinematically-derived track prediction, gating and scoring criteria are applied to greatly improve tracking efficiency and performance. Monte-Carlo testing demonstrates nearly 100 precision and strong recall in measurement assignment across a variety of multi-spacecraft formations, using both synthetic measurements and hardware-in-the-loop imagery. The algorithm can track across large data gaps and with significant measurement noise, and can cooperate with an angles-only navigation filter to improve accuracy in challenging cases.