Skip to Main Content
For systems that can be modeled using multiple models and Markovian switching probabilities, an interacting multiple model (IMM) estimator is a good approach to state estimation and mode detection. For good mode detection, it is desirable to maintain the mode-probability set so that one of the probabilities is near unity. For radar tracking a ballistic missile, the mode probability is dependent on the so-called revisit time defined as the elapsed time between each measurement. It is shown that selecting a revisit time that controls the value of a mode-switching coefficient yields acceptable performance in terms of both the mode probability and tracking error. The potential robustness of the method to tracking conditions is demonstrated through a mid-fidelity, physics-based radar simulation using targets that differ in both type and radar-to-target geometry. The results show that the proposed adaptive-revisit method performs well during boost, boost-to-coast transition, and ballistic coast, as measured by tracking error when compared to both a constant revisit interval and an ad-hoc, altitude-based revisit schedule.
Date of Conference: 2002