Comparison of data association algorithms for bearings-only multi-sensor multi-target tracking

Abstract

In multi-sensor multi-target bearings-only tracking we often see false intersections of bearings known as ghosts. When the bearing measurements from each sensor have been associated to form sequences termed threads, the problem is to associate pairs of threads to identify the true target intersections. In this paper we present two algorithms: (i) classical bayesian thread association (CBTA) and (ii) Monte Carlo thread association (MCTA), for this problem. The performance of these algorithms is compared using Monte Carlo simulations. Furthermore, we also compare their performance against the Rao-Blackwellised Monte Carlo Data Association (RBMCDA) algorithm, which uses unthreaded measurements, in order to ascertain the benefits of using thread information. Simulations show that MCTA is superior to CBTA, and that there is significant benefit in using thread information in this class of problems.