Particles filter applied in the real-time bearings-only tracking problem of a sonar target

Abstract

The development of a passive sonar tracking system in real-time needs algorithms each time more accurate and that can be able to follow rapid changes in the signal characteristics. In a bearings-only tracking problem, the bearing model is a nonlinear function of the target states. Nowadays, methods that deal with nonlinear and non-Gaussian estimation are receiving great attention. Particles filter is one of these methods. Particles filters are Sequential Monte Carlo methods that represent the required probability density function as a set of random samples. This present paper describes the particles filter application in the bearings-only tracking problem. Real data of targets are displayed in energy versus bearing graphics obtained in an operational naval environment. These data were used with the Sampling Importance Resampling algorithm (SIR). The particles filter is formulated in the Cartesian coordinates and then are transformed to modified polar coordinates. Results are compared with Kalman Extended Filter and the effectiveness and limitations of SIR algorithm for tracking target using a bearing/time record are examined.