ROBUST GAUSSIAN SUM FILTERING WITH UNKNOWN NOISE STATISTICS: APPLICATION TO TARGET TRACKING

In many real-life Bayesian estimation problems, it is appropriate to consider non-Gaussian noise distributions to model the existence of outliers, impulsive behaviors or heavy-tailed physical phenomena in the measurements. Moreover, the complete knowledge of the system dynamics uses to be limited, as well as for the process and measurement noise statistics. In this paper, we propose an adaptive recursive Gaussian sum filter that addresses the adaptive Bayesian filtering problem, tackling efficiently nonlinear behaviors while being robust to the weak knowledge of the system. The new method is based on the relationship between the measurement noise parameters and the innovations sequence, used to recursively infer the Gaussian mixture model noise parameters. Numerical results exhibit enhanced robustness against both non-Gaussian noise and unknown parameters. Simulation results are provided to show that good performance can be attained when compared to the standard known statistics case.