On Kalman Smoothing With Random Packet Loss

This correspondence studies the performance of Kalman fixed lag smoothers with random packet losses and its comparison with the Kalman filter with packet loss. In terms of estimator stability via boundedness of the expectation of the error covariance, we show that smoothing does not provide any benefit over filtering. On the other hand, it is demonstrated that using a probabilistic notion of performance, smoothing can provide significant gains when compared to Kalman filtering. An analysis of Kalman filtering using two simple retransmission schemes and its comparison with Kalman smoothing is also made.