An analysis of sampling based techniques in bearing-only tracking

Bearing only Tracking (BoT) has been of particular interest in RADAR / SONAR based passive surveillance. There are various methods and algorithms for localizing a target based on bearing measurements alone. This range from classic least square method developed by Gauss to Bayesian estimation techniques like Kalman filtering, Particle filtering etc. A review of nonlinear Bayesian filtering methods including deterministic sampling techniques like unscented filtering to random sampling techniques like Particle filters is included in this paper. Bayesian framework for nonlinear estimation is about evaluating multidimensional integrals to find out posterior and prior probability density function (pdf) of the state vector. This can be achieved via deterministic or random sampling in the state space. The main scope of this work is to compare various sampling based BoT algorithms like sigma point Alter, particle filters and cubature kalman filters in Bayesian estimation problem.