A Robust Random Matrix Prediction Model for Extended Object Rotations

This paper proposes a robust random matrix prediction model for the tracking of extended objects performing turn-rate manoeuvres. A key innovation is to derive exact solutions for two matrix-valued expectations that are currently approximated by third-order Taylor series expansions. Moreover, we prove that the use of such Taylor series expansions will lead to tracking failure when the turn-rate variance exceeds a known threshold. The exact solution can be computed in a robust and computationally efficient manner, and we demonstrate its utility in simulation compared with state-of-the-art alternative approaches.