A distributed linear equation solver for least square solutions

Xuan Wang, Jingqiu Zhou, Shaoshuai Mou, Martin J. Corless

Abstract: This paper proposes a distributed update for achieving the least square solution of linear equations by multiagent networks, in which each agent only knows part of these linear equations and is able to communicate with its nearby neighbors. It has been analytically proved that for undirected and connected network, the proposed update enables all agents to achieve an exact least square solution exponentially fast by iteratively updating its own state based on those from its neighbors. Moreover, the exponential convergence does not rely on any time-varying small step size.