Robust regression-based EKF for tracking underwater targets

In underwater target tracking applications, measurement uncertainty and inaccuracies are usually modeled as additive Gaussian noise. The Gaussian model of noise may not be appropriate in many practical systems. The non-Gaussian noise and the model non-linearity arising in a tracking system will seriously affect the tracking performance. This paper discusses one way to create a robust version of the extended Kalman filter for enhanced underwater target tracking. State estimation in the filter is done through the robust regression approach and Welsch´s proposal is used in the regression process. Monte Carlo simulation results with heavy-tailed contaminated observation noise demonstrate the robustness of the proposed estimation procedure