An innovative packet-splitting approach for kalman filtering over lossy networks

We consider the problem of state estimation over lossy networks. Although a large number of approaches have been proposed to improve the estimator´s performance, most of them demand either extra channel bandwidth or sensor energy budget. In this paper, we propose an innovative packet splitting transmission approach and derive a corresponding packet-splitting Kalman Filter (PSKF). In this scheme, one bit of each packet is diverted from quantizing the current innovation to indicate the sign of the previous innovation. We show that if converges, the expected value of the a posteriori estimate error covariance (E[P_k]) of the PSKF converges to a smaller value compared with that of modified Kalman filter in literature. Hence the proposed PSKF is able to tolerate a higher or at least equal data loss rate than the MKF. Examples are provided to illustrate the main ideas.