Event-Triggered Proportional-Integral Algorithms for Distributed Optimization (Invited Session Extended Abstract)

Abstract

In this paper, we develop event-triggered distributed optimization algorithms for undirected connected graphs based on the proportional-integral (PI) control strategy. We show that the proposed algorithms are free of Zeno behavior, and asymptotically converge to one of global minimizers, if the local cost functions are convex and differentiable. Moreover, we show that the proposed algorithms exponentially converge to the unique global minimizer if in addition, the local cost functions have locally Lipschitz gradients, and the global cost function is restricted strongly convex with respect to the global minimizer.

Publication
In 40th Chinese Control Conference

Related