Maneuvering target tracking algorithm based on marginalized Gauss-Hermite Kalman filter with sensor measurement biases

In order to address the issue of maneuvering target tracking with sensor measurement biases, a new method based on marginalised Gauss-Hermite Kalman filter (MGHKF) is proposed. Firstly, the differential measurement equation is constructed by differencing consecutive measurements to eliminate measurement biases, and the target state of adjacent moments is expanded to form an augmented state vector to match the differential measurement equation, enabling real-time filtering estimation. Meanwhile, the augmented state vector is marginalized to reduce the sampling dimension of Gaussian Hermite filtering, thereby reducing the filtering burden. Then, the interactive multiple model method is integrated into the marginalized Gaussian Hermite Kalman filtering to handle the uncertainty of motion models of maneuvering target. Based on the differential measurement equation and augmented state vector, the corresponding filtering equations are derived. Experimental results demonstrate that the proposed method can effectively eliminate sensor measurement biases and achieves significantly higher tracking accuracy compared to the traditional interactive multiple model Gaussian Hermite Kalman filtering method.