State estimation over packet dropping networks using multiple description coding

For state estimation over a communication network, efficiency and reliability of the network are critical issues. The presence of packet dropping and communication delay can greatly impair our ability to measure and predict the state of a dynamic process. In this paper, multiple description (MD) codes, a type of network source codes, are used to compensate for this effect on Kalman filtering. We consider two packet dropping models: in one model, packet dropping occurs according to an independent and identically distributed (i.i.d.) Bernoulli random process and in the other model, packet dropping is bursty and occurs according to a Markov chain. We show that MD codes greatly improve the statistical stability and performance of Kalman filter over a large set of packet loss scenarios in both cases. Our conclusions are verified by simulation results.