A novel algorithm for tracking targets using manoeuvre models

We present a novel algorithm for robust tracking of targets in challenging scenarios by modelling their high level behaviour. The particle-filter [6] based algorithm combines the benefits of variable-rate filtering [5] with a variable-dimension, manoeuvre-based representation of the target trajectory. Variable-rate filtering allows the algorithm to model manoeuvre changes at arbitrary (non-measurement) times, while manoeuvre-based modelling adds to track robustness by providing a parsimonious representation of the target trajectory. The efficiency of the particle filter has been greatly enhanced using reversible-jump Markov Chain Monte Carlo [7] moves that are applied in a Resample-Move fashion [4]. We demonstrate the performance of the algorithm on some example representative scenarios: