Matching of the estimating covariance in bearings-only tracking algorithm for moving surface targets in multiple model filters

 In this paper, an algorithm for bearings-only target tracking (BOT) problem is proposed using the resetting of the covariance matrix of the filter . Poor observability in BOT problem often leads to a bias error or even divergence of the filter. Therefore, due to non-uniqueness of the problem, the covariance value of the filter is not a sufficient condition to indicate the estimation error. In this paper, it is shown that the mutual effects between the covariance of the target range estimation and the Cramer-Rao lower error bound is similar to the stability of a Lyapunov function for detecting the divergence of the recursive filter. Then, this function is used as a correction coefficient of the estimating covariance matching. The multi-model/particle filters are also adopted for estimating the initial range of the target and calculating the Cramer-Rao lower error bound. Using the Monte Carlo simulation method, the proposed algorithm is compared with other conventional filters based on the criteria reported in the literature. Improvement of the results of the target range estimation is also shown, especially in case of observability reduction. All studies conducted in this paper are limited to the problem of surface floats tracking with passive sonar for a typical submarine.