A comparison of sample based filters and the extended Kalman filter for the bearings-only tracking problem

In this paper¹ we are concerned with the development and evaluation of effective filtering methods for the bearings-only tracking problem. For this problem we consider a stationary observer who only obtains measurements of the bearing of a moving object subject to noise. We assume a linear Gaussian evolution in the states describing the motion of the object. We develop new sample based methods for filtering for this extremely non-linear model. We compare the performance of these methods to existing sample based methods and to the extended Kalman filter. Simulated scenarios are considered for evaluating the relative efficiency of the methods considered. Finally, an actual scenario arising from recordings made on a civilian ship is considered.