复杂拓扑环境下群体博弈机制的研究

发布时间：2018-06-16 00:04

本文选题：演化博弈论 + 稀疏规则网络　；参考：《天津理工大学》2015年硕士论文

【摘要】：演化博弈论的诞生为人们描述、研究自然界对象提供了一套概念和方法,极大丰富了人们认识世界的角度,可以很好的解释一些看似矛盾、实则统一的现象,如群体间的合作与竞争等。复杂网络以其独特的抽象方式描述真实世界的复杂系统,力求体现它们的本质特征,其主要强调系统环境的拓扑特征,不仅提高了人们对现实生活中复杂系统结构的认识,更重要的是可以为描述博弈个体间的关系提供一个统一的理论框架。因此,将复杂网络研究与演化博弈理论相结合,进一步深化对空间博弈机制的理解,揭示群体间合作演化现象的本质。论文致力于研究网络结构的稀疏性、群体的多样性等对合作演化的影响,为如何促进群体间合作提供了理论性指导。论文主要研究内容和创新性成果概括如下:(1)探讨稀疏规则网络中群体合作水平的变化情况,主要考虑邻域大小、群体密度以及群体自私程度对种群合作率的影响。主要使用三种博弈模型:囚徒困境、雪堆博弈以及猎鹿博弈,仿真结果表明群体合作水平与邻域大小之间有密切关系,且一般会存在一个达到最佳合作水平的群体密度值,称为阈值。当群体密度过大时,内部意见很难一致,会影响个体之间的交流和学习,故合作率降低;尤其对于自私程度高的群体来说,合作簇难以抵挡背叛者入侵,故群体密度稍高时,合作率就会降低直至为零。(2)提出具有个体多样性的公共物品博弈模型,通过引入不同的策略传递因子,模拟有影响力个体和普通个体之间的决策过程,主要研究两种类型个体比例对合作水平的作用,并观察不同邻域大小的群体中作用程度是否相同。研究发现具有影响力个体的加入对群体合作水平有显著促进作用,而且两种个体策略学习能力相差越大,则对合作水平促进作用越大。(3)基于Java语言,设计和实现了一个简易斑图可视化软件。因考虑研究过程需要绘制大量的斑图,以方便观察群体博弈过程中合作者和背叛者的分布情况,软件基于Myeclipse开发工具主要实现了斑图的绘制及编辑功能,其简便易用的功能,不仅能够极大提高科研效率,更重要的是增强了仿真结果的直观性。
[Abstract]:The birth of evolutionary game theory provides a set of concepts and methods for people to describe and study natural objects, which greatly enriches people's understanding of the world and can explain some seemingly contradictory and unifying phenomena. Such as cooperation and competition among groups, etc. Complex networks describe complex systems in the real world in their unique abstract way and strive to embody their essential characteristics. Their main emphasis on the topological characteristics of the system environment not only improves people's understanding of the structure of complex systems in real life. More importantly, it can provide a unified theoretical framework for describing the relationship between game players. Therefore, the study of complex networks is combined with evolutionary game theory to further deepen the understanding of spatial game mechanism and reveal the essence of cooperative evolution among groups. This paper is devoted to study the influence of network structure sparsity and diversity on cooperative evolution, which provides theoretical guidance on how to promote inter-group cooperation. The main research contents and innovative results are summarized as follows: 1) the changes of the level of group cooperation in sparse rule networks are discussed, and the effects of neighborhood size, population density and group selfishness on the rate of population cooperation are considered. Three kinds of game models are mainly used: prisoner's dilemma, snowdrift game and deer hunting game. The simulation results show that there is a close relationship between the level of group cooperation and the size of the neighborhood, and there is generally a group density value that reaches the best level of cooperation. It's called a threshold. When the group density is too high, the internal opinion is difficult to agree, which will affect the communication and learning between individuals, so the cooperation rate is reduced. Especially for the group with high degree of selfishness, the cooperative cluster is difficult to resist the betrayer invasion, so the group density is higher. A game model of public goods with individual diversity is proposed. By introducing different strategy transfer factors, the decision-making process between influential individuals and ordinary individuals is simulated. This paper mainly studies the effect of the proportion of two types of individuals on the level of cooperation, and observes whether the action degree is the same in the populations with different neighborhood sizes. It is found that the addition of influential individuals can significantly promote the level of group cooperation, and the greater the learning ability of the two strategies is, the greater the contribution to the level of cooperation is based on Java language. A simple pattern visualization software is designed and implemented. In order to observe the distribution of collaborators and betrayors in the process of group game, the software mainly realizes the function of drawing and editing the pattern based on Myeclipse, which is easy to use, because of the need to draw a large number of patterns in order to observe the distribution of collaborators and betrayors in the process of group game. Not only can the efficiency of scientific research be greatly improved, but also the intuitiveness of simulation results can be enhanced.
【学位授予单位】：天津理工大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：O157.5;O225

【参考文献】