Efficient Multiagent Coordination in Dynamic Environments

Agent coordination is a fundamental task in designing and operating multiagent systems. However, in dynamically changing environments, coordination must balance proactive and reactive behaviors in order to enable efficient operations while retaining the necessary flexibility to react to unforeseen e...

Full description

Saved in:
Bibliographic Details
Published in:2011 IEEE/WIC/ACM International Conferences on Web Intelligence and Intelligent Agent Technology Vol. 2; pp. 188 - 195
Main Authors: Berndt, J. O., Herzog, O.
Format: Conference Proceeding
Language:English
Published: IEEE 01.08.2011
Subjects:
Adaptation models
adaption
Context
Decision making
dynamics
Feedback loop
Joints
Logistics
multiagent coordination
Multiagent systems
social order generation
ISBN:9781457713736, 145771373X
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:Agent coordination is a fundamental task in designing and operating multiagent systems. However, in dynamically changing environments, coordination must balance proactive and reactive behaviors in order to enable efficient operations while retaining the necessary flexibility to react to unforeseen events. This paper introduces adaptive agent relationships for coping with these contradictory requirements. In this approach, agents dynamically establish relationships which are represented as interaction patterns. On the one hand, these patterns enable efficient coordination by restricting the number of potential interaction flows to those offering the best estimated outcome. On the other hand, they can adapt to environmental changes, as the agents continuously reconsider their relationships in a feedback loop of estimated interaction flows and actually observed coordination outcomes. The paper formalizes the agent decision-making process enabling adaptive relationships and applies it to a logistics network scenario. A comparative evaluation demonstrates its ability to efficiently coordinate agent interaction in a dynamic environment.
ISBN:9781457713736
145771373X
DOI:10.1109/WI-IAT.2011.63