【作者】 申静；

【导师】 夏幼明； 孙瑜；

【作者基本信息】 云南师范大学 ， 计算机软件与理论， 2007， 硕士

【摘要】 随着计算机技术的迅猛发展,人工智能中的Agent及多Agent技术已成为研究领域的一个新热点；社会活动中经济行为的普遍化使得博弈论方法不断扩展和成熟,博弈论的应用也逐渐渗透到各不同研究领域。把博弈论的方法引入到多Agent系统中,使得Agent系统的应用更具有现实意义。本文基于博弈论和多Agent协商相结合的思想,把团体多Agent之间的合作与通信、以及不同环境下博弈策略选择的不同的方法相结合,构建了基于博弈论和多Agent协商的足球赛虚拟系统。通过对虚拟足球赛系统(Soccer Game Virtual System:SGVS)的研究,本文工作主要是以下几个方面：(1)给出Agent的相关概念和博弈论的一般要素,以及博弈要素在体育博弈中的体现,并分析了虚拟足球赛中射门博弈策略。根据虚拟足球赛SGVS系统的主要特点和功能说明在SGVS系统进入虚拟比赛之前可以对球员参数、比赛策略以及阵型等相关数据修改。(2)针对Agent资源有限性问题,引入了知识库信息和角色的概念及类型。其中,在知识库中以产生式规则的形式来存放比赛中的规则及知识,运用这些知识可以使球员之间达到很好的配合。(3)根据短周期内Agent体力消耗近似线性衰减规律,根据信息预测、角色交换及触发机制,提出了基于团体效益的团队策略和基于团体效益的角色交换算法,使得内部的协商合作获得最大效益。(4)通过模拟比赛实验分析,在虚拟比赛的实验环境中,博弈双方的球队各自采取单一的比赛策略都不是全局最优的。结果表明,运用把博弈论和多Agent协商相接合的基于团体效益的策略选择的方法是可行的、有效的,使得球队小组在不完全信息的动态博弈的环境下,能够充分利用有限的资源提高团体效益。实验结果表明,SGVS系统的构建是成功的,基于团体效用的角色交换策略有效地控制了Agent的体力并充分地利用有限的Agent资源,使整个球队球员保持较稳定的体力状态,避免了主力进攻的球员或后卫因为过度奔跑陷入疲劳状态而降低球队整体实力。Agent更多还原

【Abstract】 Along with the computer technology swift and violent development, in Artificial Intelligence Agent and the Multi-Agent technology has become the research area a new hot spot. The generalization of the economic activities in the social activity causes the game theory method unceasingly to expand and to be mature, the game theory application also seeps to different research area. Introduces the game theory method into the Multi-Agent system, enables the Agent system has the practical significance in the application. Based on the thought which unifies the game theory and the Multi-Agent consultation of this article, the cooperation and the correspondence between Multi-Agents, as well as the gambled strategy choice with different method under the different environment, this article has constructed the soccer game virtual system based on the game theory and the Multi-Agent negotiation. Through the research to the virtual soccer game system (Soccer Game Virtual System:SGVS), the main work of this article draws following several aspects:(1) Firstly gave the related concept of Agent and the game theory general essential factor, also gave the manifesting of the gambling essential factor in sports gambling, and analyzed the shoot strategy in the virtual soccer game. According to the main characteristic and the function of the virtual soccer game SGVS system, showing that before entering the virtual competition in the SGVS system can modify the player parameter, the competition strategy as well as the correlation data revision and so on.(2) In view of the Agent resources limited question, has introduced the knowledge library information and the role concept as well as the role type. Among them, deposits the rule and knowledge in the game with the production pattern rule form in the knowledge library, with the knowledge players can achieve the good coordination in the match.(3) According to the linearity weaken rule of the Agent physical strength in the short cycle, also according to the information forecast, the role exchange as well as the trigger mechanism, proposed the team strategy and the role exchange algorithm based on the team benefit, made it get the biggest benefit with the interior cooperation.(4) Through the analyzes of the competition tests, in the virtual competition experimental environment, the bilateral team respectively to adopt the sole competition strategy all is not the most superior situation in the whole. The result shown that, the method of the strategy based on the team benefit about the union of the Multi-Agent negotiation and the game theory is feasible and effective, making the team enhance the team benefit with the limited resources fully use under the complete information dynamic gambling environment.The experimental result indicated that, the constructing of SGVS system is successful, effectively controls Agents stamina and used the limited Agent resources fully, making the entire team players maintain the stable physical strength condition, has avoided the main force attack player or the fullback falls into the weary condition because excessively run to reduce the team overall strength.更多还原