Coordination for Agents with Freedom of Choice

More Info
expand_more

Abstract

Autonomous, self-interested agents have to construct plans for their activities. Together, these plans form a joint plan in the multi-agent system the agents are part of. Merging the plans of these agents is not guaranteed to be successful as their plans might conflict. Hence, we require some form of coordination that enables agents to plan for their tasks autonomously. We will present a pre-planning coordination approach. Our research focusses on coordination of tasks with preconditions and effects, which we define as instantiation-coordination. It is based on an existing approach, called plan-coordination. Verifying instantiation-coordination is coNP-complete. Ensuring instantiation-coordination is even harder: Sigma-2-p-complete. We apply our approach to a classical planning domain: Logistics. Based on these experiments, we evaluate the cost involved with our coordination approach for autonomous agents.