Decentralized estimation for linear complex networks with multi-level quantization

This paper addresses the decentralized state estimation problem for a class of discrete-time linear complex networks under communication constraints. Due to the limited communication bandwidth and radiated power, a multi-level quantization (MLQ) scheme is utilized to compress the measurement innovations transmitted over the sensor-to-estimator communication channel. In each node, a modified approximate minimum mean-square error (MMSE) estimator is constructed by sequentially fusing the quantized innovations from the corresponding sensors. The designed estimator is of a decentralized framework and relies on the state estimates and estimation error covariances from neighboring nodes. Furthermore, the quantization levels are obtained by minimizing the estimation error covariance and a sufficient condition is established to ensure the bounded estimation error covariance in each node. Finally, simulation results demonstrate the effectiveness of the proposed decentralized estimation algorithm.