Range parameterised maximum correntropy unscented Kalman filter for two dimensional angles-only target tracking problems. Issue 22 (2022)

• Record Type: Journal Article
• Authors: Urooj, Asfia; Radhakrishnan, Rahul
• Is Part Of: IFAC-PapersOnLine. Volume 55:Issue 22(2022)
• Page Start: 218
• Page End: 223
• Publication Date: 2022
• Subjects: Maximum Correntropy -- Non-Gaussian noise -- Parameterised state estimation -- Gaussian kernel; Automatic control -- Periodicals; 629.805
• DOI: 10.1016/j.ifacol.2023.03.037
• Languages: English
• ISSNs: 2405-8963

Abstract

Abstract: This paper address the angles-only target tracking (AoT) problem in two dimension where the focus is on mitigating detrimental effects of the initial range uncertainty and noise uncertainties in measurements, on estimation accuracy. To efficiently handle these uncertainties and to generate more accurate and robust state estimates, maximum correntropy unscented Kalman filter (MC-UKF) is integrated with the range parametrisation approach. As AoT problem is more challenging among the target tracking problems, the presence of non Gaussian noise in measurements combined with the initial uncertainty in range, makes it even more difficult to solve with the help of conventional estimators like the UKF. Hence we develop the range parametrised maximum correntropy unscented Kalman filter (RP-MC-UKF) and the estimation accuracy is compared with the UKF and its range parametrised version (RP-UKF). The non Gaussian noise in measurement is modelled as glint noise plus shot noise. The estimation accuracy was evaluated based on the root mean square error (RMSE) in position and the % of track-loss. It was observed that the developed RP-MC-UKF performed with significant improvement in estimation accuracy in the presence of range uncertainty as well as glint plus shot noise in angular measurements.