Target Tracking of UUV Based on Maximum Correntropy High-Order UGHF

Underwater target tracking with signal propagation delay is a major challenge. The widely-used unscented Gauss-Helmert filter (UGHF) algorithm shows good tracking performance in bearing-only filters with signal propagation delay. However, the UGHF algorithm is a nonlinear filtering method based on second-order unscented transformation (UT), which can only match the accuracy of the third-order Taylor expansion term of the Gauss-Hermite model (GHM). Moreover, UGHF is a Gaussian filter, and when the measurement noise presents non-Gaussian characteristics, the tracking accuracy will be reduced. In order to improve the tracking accuracy of underwater bearing-only target, this article proposes a high-order UGHF (HOUGHF) algorithm using high-order UT instead of second-order UT. In order to further improve the tracking accuracy of the tracking algorithm under non-Gaussian noise, in this article, the maximum correntropy criterion (MCC) is used to replace the original minimum mean square error (mmse) to update HOGGHF's posterior estimation. The maximum correntropy HOGGHF (MCHOGGHF) algorithm is proposed by using MCC based on the Gaussian kernel function. At the same time, in order to deal with the change of measurement noise, a method of adaptive Gaussian kernel width is presented. Simulation and sea trial results show that the two algorithms are superior to the UGHF algorithm and the MCHOGGH algorithm is superior to the HOGGHF algorithm in estimation accuracy under non-Gaussian noise.

Automated summary

Underwater target tracking with signal propagation delay is a challenging task. The unscented Gauss-Helmert filter (UGHF) algorithm is widely used and shows good tracking performance. However, its accuracy is limited due to its second-order unscented transformation. To improve the tracking accuracy, a high-order UGHF (HOUGHF) algorithm using high-order unscented transformation is proposed. The maximum correntropy HOGGHF (MCHOGGHF) algorithm is also proposed to handle non-Gaussian noise. Simulation and sea trial results demonstrate that the proposed algorithms outperform the UGHF algorithm and the MCHOGGH algorithm has higher estimation accuracy under non-Gaussian noise.