RD-IMM performance under sparse observations and complex maneuver profiles

Investigate the performance of the reference-driven interacting multiple model (RD-IMM) under sparse observations separated by long gaps and in scenarios involving multiple and finite-duration burns.

Background

The paper evaluates RD-IMM for detecting unannounced satellite maneuvers using space-based optical measurements collected approximately every minute from two to four sensors. Its continuous mapping from a reference coast filter’s normalized innovation squared to the coast-to-maneuver transition probability benefits from consecutive measurements, because multiple innovations can accumulate before a maneuver declaration threshold is exceeded.

The authors identify the observation and maneuver configuration as a limitation of the study. They specifically leave unresolved how RD-IMM behaves when observations are separated by long gaps, and when the target undergoes multiple maneuvers or finite-duration burns rather than the single impulsive maneuver used in the simulations.