Conditions for detectability in distributed consensus-based observer networks

TL;DR

This paper explores the fundamental conditions under which distributed consensus-based observer networks can reliably estimate system states, linking detectability, observability, and network topology to ensure stable error dynamics.

Contribution

It establishes new theoretical connections between plant detectability, node observability, and communication graph properties for stable distributed estimation.

Findings

Detectability of the plant is linked to networked observer stability.

Graph algebraic properties influence the stabilizability of error dynamics.

Conditions for output injection to ensure convergence are identified.

Abstract

The paper discusses fundamental detectability properties associated with the problem of distributed state estimation using networked observers. The main result of the paper establishes connections between detectability of the plant through measurements, observability of the node filters through interconnections, and algebraic properties of the underlying communication graph, to ensure the interconnected filtering error dynamics are stabilizable via output injection.