A GMM approximation with merge and split for nonlinear non-Gaussian tracking

In this paper, we present a recursive estimation algorithm for nonlinear non-Gaussian tracking systems based on an adaptive Gaussian mixture technique. This estimation cannot be efficiently performed for nonlinear non-Gaussian systems because of the complex representation of the state density. To alleviate this complexity, approximation techniques based on Gaussian mixtures are used. An adaptive mixture approximation method is used based on optimal minimization of a least squares error function between the true density and the corresponding approximation mixture. The state Gaussian mixture is propagated over time through prediction and update steps. Results in comparison with other methods show the efficiency of the proposed algorithm.