The Study And Design Of Multi-Agent System Negotiation Model

When faced with complicated and rapidly changing environment and unfamiliar tasks, even powerful computer system (such as computer networks, clusters, etc.) is also not satisfactory. One important reason is that different tasks require the system have different functions and different functions often require the system has a different structure (ie structure determines function.).Obviously, the conventional computer system can not meet the requirements. For a given computer system, even by some external force to change its structure, is not easy, let alone by the system its own restructuring. However, if in accordance with the Multi-Agent System model design computer systems, you can escape the above dilemma. Multi-Agent system can self-adjust the structure to be completed in a different environment with different tasks. Therefore, the computer (software or hardware) system will have more powerful problem-solving ability, especially for problems that structure is unpredictable and structure is not clear, its advantages is more prominent.For a particular problem (or task) in Multi-Agent system, different agent cooperate by independent auto-negotiated. This cooperation constitutes a specific problem solving "organization structure." For the structure formation, evolution and the whole problem solving process, there is no need to core algorithm and global control, and completely decided by Agent's independent action. Therefore, when design Multi-Agent System, we only need to focus on each group individual Agent, which greatly reduces the complexity of computer systems design.The study of traditional Multi-Agent System negotiation is always in the static, assuming complete knowledge and the premise of unlimited computing power, therefore is unable to adapt to the a complicated and changeable practical application environments.Based on traditional game theory and heuristic negotiation model, this paper proposes an improved Multi-Agent System negotiation model. Model includes two key improvements:a time-limited and incomplete information. Firstly, the negotiation model is defined by formal approach, and provides negotiation protocol in negotiation process; then defined the Multi-issues utility function; given policy rules with time limits, the rules can be used as conditions which determine whether success for negotiation process. Finally, analysis comparing several different negotiation strategies by Experimental, the results of experimental show that negotiation strategies model can reach negotiation in limited time under the incomplete information, and improve the effectiveness of negotiation, which lay the concrete for its implementation as well.