Covariance Consistency of Tracking Filters in ATC Systems

摘要

This paper reviews the techniques used by various filters to ensure covariance consistency under non-linear tracking situations. In addition to state estimates, filters such as the Kalman filter provide an estimation covariance matrix, which quantifies the accuracy of the state estimate. In applications such as air traffic control, the state estimation covariance is used to predict target future position region. In the polar measurement situation, the original Kalman filter is usually replaced by an extended Kalman filter or a converted measurement Kalman filter, but the consistency of the state estimation covariance is no longer guaranteed. This paper compares the estimation covariance consistency of the classic converted measurement Kalman filter, the modified unbiased converted measurement Kalman filter, and the particle filter. From the simulation results of an aircraft tracking scenario, all three filters have good covariance consistency under small azimuth noise. For large azimuth noise, the particle filter has the best consistency, while the classic converted measurement Kalman filter has very poor consistency.