Optimal passive tracking of ground targets

The performance of the EKF (extended Kalman filter) as a state estimator in a restricted passive tracking problem is explored. The performance of the EKF was compared to an approximate optimal minimum variance estimate. The EKF was found to have poorer state estimate convergence, and poorer agreement between the estimator predicted error variance and the actual error variance. Viewing the EKF as another approximate Bayes estimator pointed out that the EKF deficiencies are partly as a result of the mismatch between the EKF assumption of Gaussian update densities and the true nonGaussian measurement update density