In this paper, a hybrid Monte Carlo Gauss mixture Kalman filter is proposed for the continuous orbit estimation problem. Specifically, the graphic processing unit (GPU) aided Monte Carlo method is used to propagate the uncertainty of the estimation when the observation is not available and the Gauss mixture Kalman filter is used to update the estimation when the observation sequences are available. A typical space object tracking problem using the ground radar is used to test the performance of the proposed algorithm. The performance of the proposed algorithm is compared with the popular cubature Kalman filter (CKF). The simulation results show that the ordinary CKF diverges in 5 observation periods. In contrast, the proposed hybrid Monte Carlo Gauss mixture Kalman filter achieves satisfactory performance in all observation periods. In addition, by using the GPU, the computational time is over 100 times less than that using the conventional central processing unit (CPU).

Date Published: 13 May 2016
PDF: 12 pages
Proc. SPIE 9838, Sensors and Systems for Space Applications IX, 983803 (13 May 2016); doi: 10.1117/12.2224966

Published in SPIE Proceedings Vol. 9838:
Sensors and Systems for Space Applications IX
Khanh D. Pham; Genshe Chen, Editor(s)