Convergence of the Distributed SG Algorithm Under Cooperative Excitation Condition

TL;DR

This paper introduces a distributed stochastic gradient algorithm for sensor networks that guarantees convergence under a cooperative excitation condition, even without traditional independence assumptions, and demonstrates its effectiveness through simulations.

Contribution

It proposes a novel distributed SG algorithm with a cooperative excitation condition that ensures convergence without relying on independence or stationarity of regressors.

Findings

Convergence is achieved under the cooperative excitation condition.

The convergence rate of the algorithm is established.

Sensors can collaboratively estimate parameters even if individual sensors cannot.

Abstract

In this paper, a distributed stochastic gradient (SG) algorithm is proposed where the estimators are aimed to collectively estimate an unknown time-invariant parameter from a set of noisy measurements obtained by distributed sensors. The proposed distributed SG algorithm combines the consensus strategy of the estimation of neighbors with the diffusion of regression vectors. A cooperative excitation condition is introduced, under which the convergence of the distributed SG algorithm can be obtained without relying on the independency and stationarity assumptions of regression vectors which are commonly used in existing literature. Furthermore, the convergence rate of the algorithm can be established. Finally, we show that all sensors can cooperate to fulfill the estimation task even though any individual sensor can not by a simulation example.