An Integrated Multi-Agent Coordination Including Planning, Scheduling, and Task Execution

Traditional planning has been explored in depth. Most research is concentrated at a high level, e.g., developing coordination interaction protocols to be imposed on agents. There has been less concern about how the internal task structures of individual agents affect these higher-level coordination behaviors. When involving multiple participants, collaborative planning has been facing problems like uncertainty, contingencies, and evaluating worth-oriented goals. In particular, agent planning, scheduling and execution are inextricably linked to coordination behaviors. This paper presents extensions and restrictions, as extended hierarchical task networks (EHTN), to the expressiveness of traditional plan and schedule representations that allow the formal definition of the multi-agent coordination problem. Based on EHTNs, this paper discusses some open issues in multi-agent coordination (e.g. what to coordinate among agents, how much information to be exchanged, how to evaluate a planning approach) and proposes a generalized solution towards successful distributed goal achievement by analyzing the task performance of participating agents.