A comparative study of nonlinear filters for target tracking in mixed coordinates

Abstract

The measurement model nonlinearity is a major challenge in target tracking. This paper presents a comparative performance study of seven nonlinear filters in handling the measurement model nonlinearity. They are: the extended Kalman filter, the unscented filter, the second order divided-differences filter, the Gauss-Hermite quadrature filter, the two-step Kalman filter, the Gaussian particle filter, and the linear minimum mean-square error tracking filter with polar measurements. Comprehensive performance evaluation and comparison of all of the above mainstream nonlinear filters over the same tracking scenarios are conducted via Monte Carlo simulation. The results can facilitate the choice and design of nonlinear tracking filters in mixed coordinates.