AAAI Conference 1999 Conference Paper
Learning Quantitative Knowledge for Multiagent Coordination
- David Jensen
- Michael Atighetchi
- Régis Vincent
- Victor Lesser
- University of Massachusetts at Amherst
A central challenge of multiagent coordination is reasoning about howthe actions of one agent affect the actions of another. Knowledge of these interrelationships can help coordinate agents -- preventing conflicts and exploiting beneficial relationships among actions. Weexplore three interlocking methods that learn quantitative knowledge of such non-local effects in T/EMS, a well-developed frameworkfor multiagent coordination. Thesurprising simplicity and effectiveness of these methods demonstrates howagents can learn domain-specificknowledge quickly, extendingthe utility of coordination frameworks that explicitly represent coordination knowledge.