Distributed Sensor Fusion Using Dynamic Consensus
Abstract
This work is an extension to a companion paper describing dynamicconsensus of networked agents, and shows how those results can beapplied to obtain a distributed Kalman filter for a network ofagents. The underlying mechanism is the Laplacian consensusdynamics, which utilizes peer-to-peer interactions, and is veryrobust to time-varying topology changes. The performance of thedistributed estimator is analyzable in terms of properties of thenetwork, and degrades gracefully with decreasing networkperformance. Further, this mechanism can be used in a nativelyasynchronous mode, making it appropriate for real-worldpeer-to-peer networks.
