Distributed Online Optimization With Communication And Feedback Delays

This paper studies the application of distributed online convex optimization in multi-agent systems under communication and feedback delays. Comprising a network topology, multiple sub-agents engage in the estimation of the global optimal strategy through the exchange of their individual information. We augment the communication network with virtual delay nodes. Feedback delays are also incorporated into the network. In the analysis of the algorithm's feasibility, consideration is given to the regret. We show that the regret has a sublinear bound when the communication delay is bounded and the delayed feedback can be received.