MEE-Based Adaptive State Estimator for Non-Gaussian Radar Measurement

In radar, particularly in nonlinear scenarios, the estimation of targets' kinematic parameters and track updating is carried out using nonlinear extensions of the Kalman filter (KF). The two most widely employed extensions of the KF are the extended Kalman filter (EKF) and the unscented Kalman filter (UKF). Despite claims that EKF and UKF better handle nonlin-earity, they rely on approximation methods involving Jacobian and sigma points, respectively. Consequently, the approximation of nonlinearity raises concerns about the suitability of EKF and UKF when dealing with highly nonlinear models. Moreover, they heavily depend on the Gaussianity assumption and employ the standard minimal mean square error (MMSE) criterion for filtering. In the literature, to address non-Gaussianity, the minimum error entropy (MEE)-based EKF has been extensively studied, purportedly offering superior performance over MMSE in handling non-Gaussianity. Nevertheless, like EKF, MEE-based EKF is also contingent on precise knowledge of the nature of nonlinearity. In addressing the aforementioned issues in EKF and MEE-based EKF, this work proposes a unified solution to handle both nonlinearity and non-Gaussianity in the standard state estimation problem. The proposed approach suggests combining the kernel recursive MEE (KRMEE)-based adaptive filter with the MEE-based EKF, resulting in a filter named EKF-KRMEE. The EKF-KRMEE filter adaptively estimates nonlinearity with each new radar measurement, and the evoked MEE criterion addresses the non-Gaussianity of the measurements. Computer simulations are employed to demonstrate the superiority of the proposed EKF-KRMEE filter over conventional EKF and MEE-based EKF.