On Cooperation in a Multi-Entity Model

The problem of modeling a variety of domains within the framework of one general scheme is of central importance in AI. This paper presents the MultiEntity model for multi-agent planning (MAP). This is a model in which a wide variety of scenarios can be represented, while many basic questions in MAP can be naturally represented and relatively efficiently computed. We use this model in order to study issues in MAP that we think are fundamental features of cooperation. These issues include: Achievement of cooperative goals, achievement of cooperative goals in the presence of failures, stable deals among agents (agents might not accept solutions that are irrational to them), tolerating agents that may deviate from a plan in order to improve their situation, and tolerating agents that might deviate in unknown ways from a plan (because they are liars, or because their plan has changed).

• Type: Conference
• Year: 1989
• Where: IJCAI
• Authors: Moshe Tennenholtz, Yoram Moses
• Keywords: Central Importance, Cooperative Goals, IJCAI 1989, IJCAI 2007, Model