Reflections on the Nature of Multi-Agent Coordination and Its Implications for an Agent Architecture

  • Authors:

  • Victor R. Lesser

  • Affiliations:

  • Computer Science Department, University of Massachusetts, Amherst, MA

  • Venue:
  • Autonomous Agents and Multi-Agent Systems
  • Year:
  • 1998

Quantified Score

Hi-index 0.00

Visualization

Abstract

The development of enabling infrastructure for the next generationof multi-agent systems consisting of large numbers of agents and operatingin open environments is one of the key challenges for the multi-agentcommunity.Current infrastructure support does not materially assist in thedevelopment of sophisticated agent coordination strategies. It is the needfor and the development of such a high-level support structure that will bethe focus of this paper. A domain-independent (generic) agent architectureis proposed that wraps around an agent's problem-solving component inorder to make problem solving responsive to real-time constraints, availablenetwork resources, and the need to coordinate—both in the large andsmall—with problem-solving activities of other agents. Thisarchitecture contains five components, local agent scheduling, multi-agentcoordination, organizational design, detection and diagnosis, and on-linelearning, that are designed to interact so that a range of differentsituation-specific coordination strategies can be implemented and adapted asthe situation evolves. The presentation of this architecture is followed bya more detailed discussion on the interaction among these components and theresearch questions that need to be answered to understand theappropriateness of this architecture for the next generation of multi-agent systems.