Student T-Based Maximum Correntropy Unscented Kalman Filter for UAV Target Tracking

Abstract

Considering that the Student T distribution has heavy-tailed non-Gaussian property, the heavy-tailed non-Gaussian noises induced by strong maneuvering target are modeled as the Student T distribution, and a cost function based on the Student T distribution as the kernel function is designed. On this basis, the Student T-based Maximum Correntropy Unscented Kalman filter (TMCUKF) is proposed based on the designed Student T distribution cost function together with the maximum correntropy criterion. In addition, the convergence condition and proof of the proposed method are also given. This algorithm has strong suppression ability to the heavy-tailed non-Gaussian noise, and has the ability to improve the tracking accuracy.