Distributed Online Convex Optimization With Adaptive Event-triggered Scheme

This paper is concerned with the distributed online convex optimization problem in which a series of agents try to track the minimizer of a global convex function. A consensus then adaptive with gradient exchange (CTAGE) algorithm is proposed where the gradient information exchanged among agents is adopted to facilitate reaching consensus. To alleviate the communication overhead among agents, an adaptive event-triggered scheme (AETS) is adopted, which can dynamically adjust the threshold. Specifically, the threshold would hold when there are consecutive triggers, and it would decrease while the triggering condition is not satisfied for several steps. Then, theoretical results indicate that dynamic regret of CTAGE can reach sublinear upper bound. Finally, a target tracking example is utilized to demonstrate the usefulness of the proposed algorithm.