Consensus-Based Distributed Parameter Estimation with Asymmetric Communications

Abstract

This chapter focuses on exploiting the sensor fusion capability for situation monitoring applications over a kind of relay assisted sensor networks consisting of multiple kinds of SNs and RNs. SNs implement assimilation of new measurement and cooperation with other nodes. While for RNs, the main role is to aggregate their neighboring data. Moreover, SNs have different sensing modalities, which can only measure a part of the target parameter vector for situation monitoring. We propose a distributed consensus based unbiased estimation (DCUE) algorithm for this kind of sensor network. Different from existing algorithms, the DCUE algorithm explicitly takes the heterogeneity of responsibilities between SNs and RNs into account. By using algebraic graph theory in conjunction with Markov chain approach, we demonstrate how the distributed estimation method can be transformed to circumvent the challenges arisen from the heterogeneity. We analyze the performance of asymptotic unbiasedness and consistency of the DCUE algorithm in the presence of asymmetric communication, i.e., when a node can receive information from another node but not vice versa. Furthermore, a quantitative bound on the rate of convergence is established. Finally, simulation results are provided to validate the effectiveness of the DCUE algorithm. It is also demonstrated that the presence of RNs does contribute to the estimation accuracy and convergence rate compared with the homogeneous networks.