网络博弈模型中均衡解的预测方法研究

发布时间：2018-07-14 08:24

【摘要】：网络中拥有决策权的实体各自通过改变策略优化自己的目标,并影响其他个体的策略选择。为研究该场景下的资源分配问题,一种研究方式是将其建模为博弈模型并研究其均衡点。本文通过研究不同场景下的均衡预测算法,完成了如下工作:发展了两种服务器选择问题场景下的服务器选择机制;发展了特定网络场景下纳什均衡的求解算法。第一章,本文阐述了均衡预测的相关研究以及网络服务提供商和内容提供商的关系,介绍了服务器选择问题的研究现状,提出本文的研究意义。第二章,在研究用户服务器为物理网络的边缘路由器的服务器选择问题时,本文利用Wardrop均衡预测算法发展了一套分布式的服务器选择机制,并且在博弈场景和动态场景中的仿真实验当中测试其性能。实验结果显示,相比于对比方案,该选择机制在链路利用率较大的场景中可以获得更低的用户延迟。第三章,在研究用户节点为独立用户的服务器选择问题场景时,本文利用相关均衡的预测算法发展了一套该场景下的服务器选择机制,并研究在部署该方案的前提下利用遗传算法框架改进网络服务提供商的流量工程方案以优化用户延迟。实验证明相对于对比方案,遗传算法得到的方案能够降低用户的总延迟。第四章,本文针对网络服务提供商和内容提供商参与资源竞争的场景,提出了一种该场景下基于松弛算法的Nash均衡点计算方案。仿真实验说明,该方案仅适用于部分场景中。此后利用仿真试验考察了该特定场景下Nash均衡点性能,结果显示在链路占用率较高时,均衡点下二者收益相比放任其各自优化更高。第五章,对全文进行总结,并提出了下一步的建议。
[Abstract]:The entities with decision-making power in the network optimize their goals by changing their strategies and influence other individuals' policy choices. In order to study the resource allocation problem in this scenario, one way is to model it as a game model and study its equilibrium point. By studying the algorithms of equilibrium prediction in different scenarios, this paper has accomplished the following work: developing two kinds of server selection mechanism in the server selection problem scenario and developing the algorithm of solving Nash equilibrium in a specific network scenario. In the first chapter, the research of equilibrium prediction and the relationship between network service provider and content provider are introduced, and the research status of server selection is introduced, and the significance of this paper is put forward. In the second chapter, we develop a distributed server selection mechanism by using Wardrop equilibrium prediction algorithm when we study the server selection problem in which the user server is the edge router of the physical network. And test its performance in the simulation experiment of game scene and dynamic scene. The experimental results show that compared with the comparison scheme, the proposed scheme can obtain lower user latency in scenarios with high link utilization. In the third chapter, when we study the server selection scenario where the user nodes are independent users, this paper develops a server selection mechanism in this scenario by using the prediction algorithm of correlation equilibrium. In order to optimize the user delay, the genetic algorithm (GA) framework is used to improve the traffic engineering scheme of the network service provider (NSP) in order to optimize the user delay. Experimental results show that the proposed scheme can reduce the total user delay compared with the contrast scheme. In chapter 4, aiming at the scenario of network service provider and content provider participating in resource competition, we propose a scheme of Nash equilibrium point calculation based on relaxation algorithm in this scenario. The simulation results show that the scheme is only suitable for some scenarios. After that, the performance of Nash equilibrium points in this particular scenario is investigated by simulation experiments. The results show that when the link occupancy is high, the income of the two equilibrium points is higher than that of their respective optimization. The fifth chapter summarizes the full text and puts forward suggestions for the next step.
【学位授予单位】：电子科技大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP393.0

【相似文献】