Particle algorithms for filtering in high dimensional state spaces: A case study in group object tracking

We briefly present the current state-of-the-art approaches for group and extended object tracking with an emphasis on particle methods which have high potential to handle complex structured scenarios. The big dimensionality attributed to the group tracking problem poses a major difficulty to particle filters (PFs). This in turn has motivated researchers to devise many alternatives and variants over the past decade. In this work, we corroborate and extend a single promising direction for alleviating the dimensionality problem. Our derived scheme endows a recently introduced Markov chain Monte Carlo (MCMC) PF algorithm with an improved proposal distribution. We demonstrate the performance of our approach using a nonlinear system with up to 40 states.