Parameter Estimation of Radar Noise Model for Terrain Referenced Navigation Using a New EM Initialization Method

In this paper, we present an expectation-maximization (EM) initialization method for estimation of unknown parameters in a radar noise model. We deal with radar based terrain referenced navigation (TRN) over vegetated areas, where the radar noise was modeled as a double-mode Gaussian mixture with two unknowns. The known parameters are used for EM initialization along with innovation samples. Through TRN simulation, it is shown that the proposed EM initialization method outperforms random or k-means initialization methods in terms of both the parameter estimation error and the vehicle position error.