| As multi-agent systems are involved in more and more complicated applications, negotiation, a technique that can be used effectively to coordinate the behavior of agents, becomes increasingly important and difficult.; This dissertation addresses a number of problems for sophisticated negotiation in multi-agent systems. The agents are large-grained and complex with multiple goals and tasks. The organizational relationships among agents are heterogeneous, and the agents negotiate over multiple topics. Finally, the negotiation process is tightly interleaved with agents' controlling, scheduling and planning processes.; I identify the following problems: negotiation over multiple inter-related issues; negotiation in a complex organizational context; negotiation as a distributed search process. I then attack these problems from several directions. In negotiation decision process, I use the motivational quantities ( MQ) framework to support reasoning based on organizational concerns; I build a partial order scheduler that allows the agent to evaluate the flexibility and the feasibility of multi-linked negotiation issues. I also introduce the multi-leveled negotiation framework, which allows the negotiation to be performed which allow the agents negotiate in a multi-dimensional space.; I implement these approaches and validate them through experimental work. I show that efficient negotiation techniques improve agents' performance significantly. I also outline directions for improving and extending our approaches. |
