Coupled Tracker and Identification Algorithms

"Toyon Research proposes to develop a particle filter-based Bayesian state estimation algorithm. Motivated by the increasing availability of high-range-resolution radar (HRRR) data and the benefits of incorporating "feature" information into trackingalgorithms, we will develop a particle filtering algorithm which utilizes feature information in HRRR data for coupled tracking and identification. Several attributes make the particle filter an attractive approach for the joint tracking andidentification of multiple ground targets. Specifically, (1) Particle filters do not rely on linearity or require Gaussian likelihoods and naturally handle non-Gaussian likelihood functions or those available through table look-up only. (2) Formulti-target tracking, no propagation of association hypotheses is required to obtain the optimal Bayesian solution. It is sufficient to generate hypotheses at the current frame only. (3) The state space geometry is no impediment to implementing theparticle filter; thus, road-constrained ground tracking is easily accomplished without ad-hoc approximations. The resulting multi-target joint tracking and identification algorithm will be evaluated using an event-based simulation called SLAMEM.Implementing the filter within the simulation will allow for an analysis and demonstration of the filter's required run-time for cases with large numbers of targets. The successful completion of this research shall result