Robust Space Object Track Initiation Based on Constrained Optimization and the Unscented Transform

In space situational awareness (SSA), track initiation, which establishes tracks from a few initial observations, is a key technique that governs overall system performance. This paper proposes a constraint-based optimization technique for robust and accurate initial state estimation using only two radar position observations. The proposed technique minimizes the Mahalanobis distance while utilizing the quasi-circular orbit characteristics of low-earth orbit objects as physical constraints on velocity direction and magnitude. It moreover enhances the stability of the subsequent filter by quantitatively modeling the inherent uncertainty of the constraints and incorporating it into the initial error covariance. Through Monte Carlo simulations, the proposed technique was demonstrated to significantly reduce the track loss rate and improve tracking accuracy when compared with conventional two-point initiation methods.