Tracking correlated equilibria in clustered multi-agent networks via adaptive filtering algorithms

We present a decentralized adaptive filtering algorithm in a clustered network of agents. Agents receive payoffs partly due to performing localized tasks in clusters and partly due to strategic interaction with agents outside clusters. Each agent is only aware of the actions of others within its cluster and is oblivious to the actions, or even existence, of agents outside the cluster. We show that the global behavior of the network converges to the set of correlated e-equilibria if the agents follow the proposed algorithm. Thus simple behavior by individual agents can result in sophisticated global behavior.