Commitment-Based Multiagent Decision Making

In a cooperative system, multiple dynamic entities work together and share their resources to achieve common goals, while simultaneously pursuing their individual goals. In real-world business environments, participants interact by exchanging goods and providing services for each other. In seeking and providing services, the participants form associations, make promises, commit to levels of functionality and quality, satisfy what they promised, and attempt to achieve their intended goals. We believe that in an environment where software agents are the participants, it is the binary relationship of commitment that associates the agents with one another and represents multiagent interactions. Commitments can characterize-from an external viewpoint-not only the interactions between the agents, but also the overall multiagent system behavior. Commitments are a good basis for publicly describing and monitoring the interactions among agents in a multiagent system, but they do not specify how an agent privately should manage its commitments. A belief-desire-intention (BDI) architecture is a good metaphor for building a single agent, but not for building a multiagent system. Our goal is to unify these formally. Our approach to this problem is to use the agent\u27s beliefs, desires, and intentions to make decisions about commitments. Moreover, a commitment-driven decision theory can be utilized to expressively model a cooperative multiagent environment. In this dissertation, we have blended the two very robust and widely accepted theoretical frameworks (BDICTL* and commitments) to model a service-oriented multiagent system, thereby formalizing commitments in terms of agent\u27s beliefs, desires, and intentions. We have provided the basic framework using which an agent can decide rationally when to accept, abandon, cancel, or devote resources to a commitment. The agent can also decide rationally in which order to satisfy its commitments