State dependent multiple model-based particle filtering for ballistic missile tracking in a low-observable environment

This paper proposes a new method for tracking the whole trajectory of a ballistic missile (BM), in a low-observable environment with 'imperfect' sensor measurement incorporating both miss detection and false alarms. A hybrid system with state dependent transition probabilities is proposed where multiple state models represent the ballistic missile movement during different phases; and domain knowledge is exploited to model the transition probabilities between different flight phases in a state-dependent way. The random finite set (RFS) is adopted to model radar sensor measurements which include both miss detection and false alarms. Based on the proposed hybrid modeling system and the RFS represented sensor measurements, a state dependent interacting multiple model particle filtering method integrated with a generalized measurement likelihood function is developed for the BM tracking. Comprehensive simulation studies show that the proposed method outperforms the traditional ones for the BM tracking, with more accurate estimations of flight mode probabilities, positions and velocities.