Evaluation of adaptive Kalman filtering methods for target tracking applications

Abstract

Four methods of process noise covariance tuning in a Kalman filter are evaluated. The methods studied are based on four approaches: a model based heuristic approach, a method based on a piecewise constant acceleration model, a suboptimal method based on optimization of the state estimation performance of the Kalman filter, and a covariance matching technique implemented using fuzzy logic. The methods are described in the order of increasing computational complexity. The performance of the Kalman filter incorporating the four methods for tuning is compared for simulated data of a target and real data of a typical launch vehicle. The tracking performance of all the methods is almost similar, and hence the choice of a method for a particular application would depend on the resources available. The algorithms are implemented in MATLAB on a Windows NT workstation.