Study On The Technology Of Agent-Based Auto-Negotiation

Agent-based auto-negotiation technology is the innovation of traditional negotiation technology and promotes the development of the DAI in terms of theory and practice. Multi-issue negotiation between agents is a complicated course in which negotiating agents mutually exchange offers. It is very important for agents to reach an agreement and maximize their own utility under the situation that the environment and opponent information are not completely known.This thesis analyzes and discusses the current agent-based auto-negotiation technology and explicates the main ideas, developments, advantages and disadvantages of game theory based, heuristic based and argumentation based negotiation technology. Then the models of bilateral multi-issue negotiation and multilateral multi-issue negotiation are explored respectively.An agent-based bilateral multi-issue negotiation model with incomplete information and self-interested agent is proposed. Without mediators participating, it avoids the unfair or failure of negotiation caused by the bias or breakdown of mediator effectively. This model allows agents to reach a Pareto-efficient agreement through inter-detection. It will decrease the deadlock in negotiations and make the two sides gained the satisfaction results with low computational cost. Experimental results show the performance of the improved model.A GA-based agent multilateral multi-issue negotiation model with limited computing resources and incomplete knowledge is proposed. It dynamically estimates opponent's preference according to the change of the opponent's offer in the process of negotiation and then improves the sense of the fitness function. It can accelerate agents to reach a Pareto-efficient agreement and obtain a "win-win" outcome finally. The performance of the improved model was proved by experiments.