An Architecture for Adaptive Coordination of Heterogeneous Agents

We present a novel architecture to design multi–agent systems. Each agent is described by features that can dynamically change, while it is participating to the common activity. Basing on these features, a coordination system assigns agents to jobs and schemas, i.e., complex activities which require the collaboration of more than one agent. The coordination system can be distributed on the agent network, according to their communication possibilities, which may dynamically change. The whole architecture is designed to host learning and adaptation mechanisms, that ensure robustness of the multi-agent system operating in dynamical environments.