State estimation over unreliable network

We consider the problem of state estimation of a discrete time process over an unreliable communication network. Most previous work in literature focus on either quantization effect or packet drops or delays, with a few work considering those uncertainties simultaneously. In this paper, we first present two quantized filter design methods based on a logarithmic quantizer. The first method characterizes the quantization error in a deterministic way and presents a worst-case design while the second one models the quantization error as a random variable and gives an average design. The effects of packet drops and delays on the quantized filters are then evaluated, leading to a single unified framework for simultaneously dealing with quantization, packet drops and delays. Sufficient conditions on the stability of estimation error covariance matrix are provided.