A modified marginalized Kalman filter for maneuvering target tracking

摘要

This paper uses a new nonlinear Kalman filter algorithm in maneuvering target tracking. Marginalized KalmanFilter (MKF) algorithm expresses the system's nonlinear measurement equation as weighted sum of Hermite polynomials, then models the prior distribution of the weighted matrix as a Gaussian process. After that, calculates the weighted matrix's posteriori distribution and removes the influence of the weighted matrix by calculating its integration. As a result, the closed-form of the system's state and its covariance are available. To improve the stability of the MKF algorithm, Strong Tracking Filter concept is brought in by using a fading factor in the MKF algorithm. It can decrease the influence of previous filtering step on the current step. The Strong Tracking Marginalized Kaiman Filter (STMKF) algorithm has a adaptability to the sudden maneuver of the target. Using STMKF with UKF, CKF, and standard MKF in high maneuvering target tracking problem, the result shows that the STMKF has a better stability and a higher accuracy.