Annotating Cooperative Plans with Trusted Agents

摘要

Cooperation is the single most fundamental characteristic of multi-agent systems, and much work has been done on studying the various aspects involved, from general models of the overall structure of cooperation to detailed analyses of specific components. In our work, we aim to do both ― we provide a general model and instantiate each stage in that model. We take the notions of trust and motivation to be fundamental to engendering successful cooperation between autonomous entities, and our model of cooperation accounts for the important roles played by these concepts. This paper focuses in particular on the details of how, based on trust, an agent chooses and keeps track of which agents it may use to assist in the performance of actions that make up a multi-agent plan, and how that information can be used in actually soliciting the assistance.