Low-observable maritime tracking using Monte Carlo Markov chain methods

Kirubarajan and Bar-Shalom (see IEEE Transactions on Aerospace and Electronic Systems, 1996) addressed an example tracking problem which is illustrated. At 30 second intervals, bearing measurements are made from a target which is assumed to be travelling in a straight line at constant (unknown) speed; these bearing measurements are subject to errors with a Gaussian distribution. The bearing measurements are illustrated. The task is to estimate the target track on the basis of the bearing measurements alone. Observability is already compromised in this scenario because of the absence of any range measurements, and the use of only a single bearing sensor. The problem is compounded by the fact that the target yields a very low signal/noise ratio, so that in each sonar scan, the bearing contact (if any) deriving from the target is mixed with numerous false alarms due to noise. We present our interim findings in tackling the same problem from a Bayesian statistical standpoint, using Monte Carlo Markov chain methods