Dynamic coalition structure generation for autonomous connected vehicles

In this paper, we present a decentralized, dynamic Coalition Structure Generation (CSG) method for autonomous connected vehicle convoying. A set of vehicles is partitioned into coalitions, each managed by a leader. In the two-phased approach, leaders first communicate with each other to determine whether a coalition merge is beneficial. Then, they decide whether to grant their members' requests to leave, join or form new coalitions. We have implemented our algorithms in MATISSE a large-scale multi-agent based traffic simulation system. Our initial results show an improvement over recent published work on dynamic CSG for connected vehicles.