ABSTRACT
In a multi-agent system, information may be asymmetric in the sense that each agent only knows its own model, cost, and future action, but not others’. Information asymmetry may lead to undesired consequences of the system. This chapter investigates methods to resolve conflicts and facilitate cooperation in multi-agent systems under information asymmetry. In particular, two kinds of problems were investigated: control and learning strategies in quadratic games with information asymmetry, and communication-enabled conflict resolution strategies for systems with multiple Nash Equilibria. The first problem arises naturally during multi-agent collaboration. We discuss a method to compensate the uncertainty of others’ behaviors in order to ensure stability of the multi-agent system. The second problem arises in multi-vehicle navigation in unmanaged intersections. A communication protocol and strategy is discussed to build a consensus among all agents.
