Multiple Space Object Tracking Under Epistemic Uncertainty

A major challenge in most space object detection and tracking algorithms is that the lack of perfect knowledge to characterize some uncertainty sources in the system, i.e., the dynamics and observation of the target, in a probabilistic fashion. In this paper, we propose a robust multiple space object detection and tracking method by leveraging the concept of outer probability measures. The Possibility Generalized Multi-Bernoulli (GLMB) filter is employed to account for the absence of precise measurements and dynamic model parameters. In order to initiate the orbital state of a new space object in the context of Possibility GLMB, the Possibilistic Admissible Region (PAR) method is introduced to provide a credibilistic initial orbit determination based on imperfect information. The developed PAR-GLMB scheme provides an improved robustness in the situation of partial knowledge about the system. The feature of the proposed PAR-GLMB method is illustrated using a simulated Geostationary space object tracking scenario.