复杂网络中的对称性破缺博弈与级联故障

发布时间：2018-06-10 02:49

本文选题：复杂网络 + 节点交互对称性　；参考：《中国科学技术大学》2015年硕士论文

【摘要】：复杂网络常用来抽象表示某个实际系统。网络中的节点代表实际系统中的个体对象,连边代表个体对象之间的联系。在许多真实系统中,节点个体都是具有行为决策能力的主观个体。当节点间存在交互关系时,节点会采取利己行为,使交互关系向有利于自己的方向发展,打破了交互对称性。如在Mesh无线网中,各个路由器之间存在着数据包的互相传递,当其中的某个路由器节点发生拥塞时,该节点往往希望自身承担的负载可以重分发到网络中的其它节点上,同时减少来自其它节点重分发的数据包或以该节点为中转的数据包,以此来保持该节点不会效率下降或宕机。当节点过载或故障时,其上的负载会重分配到其它节点上,引起网络过载级联。然而现有级联故障的研究集中于从纯粹的拓扑结构角度考虑流量变化与拥塞控制,并没有考虑节点间交互关系的作用。基于此,本课题组构建表征节点间交互对称性破缺这一动力学行为的对应模型,并在级联故障中,研究其对网络鲁棒性的影响。本文首先揭示了真实网络系统中具有行为决策能力的节点间交互行为的对称性破缺现象；根据节点自主性、唯理性以及交互过程中的利己行为,构建对称性破缺博弈理论来表征节点间的交互对称性破缺这一动力学过程。理论上分析了模型的合理性以及影响对称性破缺的主要因素。为了探究不对称因子对博弈结果的影响,分别在ER随机网络、NW小世界网络以及BA无尺度网络上进行博弈仿真实验。实验结果验证了该博弈理论的正确性以及对称性破缺结果与不对称因子的关系。该理论模型为研究节点间交互动力学提供了理论指导。之后,本文研究了负载重分配过程中,节点间交互对称性破缺与级联传播的关系,探索了网络中的博弈动力学与级联动力学间内在的联系。我们构建了对称性破缺博弈级联模型,分别在BA网络与AS自治网络上进行级联仿真。通过调节博弈中的初始负载不对称因子与策略收益不对称因子,观察在不同攻击策略下,网络最大连通尺度与级联持续时间的变化。最后,根据所观察的级联演化的时间特性,设计了通过实时调节对称性破缺博弈模型中的参数来降低级联损失的方法。这一成果为节点具有行为决策能力的网络中由级联故障引起的重大事故提供了新的决策依据与解决方案。
[Abstract]:Complex networks are often used to abstractly represent a real system. The nodes in the network represent the individual objects in the actual system, and the connected edges represent the relationships between the individual objects. In many real systems, nodal individuals are subjective individuals with behavioral decision-making ability. When there is interaction between nodes, the node will take self-interest behavior, which makes the interaction develop in the direction that is beneficial to itself and break the symmetry of interaction. For example, in mesh wireless network, there is a transmission of packets between routers. When one of the router nodes is congested, the node often wants its load to be redistributed to other nodes in the network. At the same time, the number of packets redistributed from other nodes or transited by the node is reduced to keep the node from decreasing efficiency or downtime. When the node overload or failure, the load on the node will be redistributed to other nodes, causing network overload cascade. However, the existing research on cascading faults focuses on considering the traffic variation and congestion control from the perspective of pure topology, and does not consider the interaction between nodes. Based on this, our team constructed a corresponding model to characterize the dynamic behavior of the symmetries breaking between nodes, and in the cascade fault, First of all, this paper reveals the symmetry breaking phenomenon of the interaction behavior between nodes with behavioral decision ability in real network system, according to node autonomy, rationality and self-interest behavior in the process of interaction. The symmetry breaking game theory is constructed to characterize the dynamic process of the interactive symmetry breaking between nodes. The rationality of the model and the main factors affecting symmetry breaking are analyzed theoretically. In order to explore the influence of asymmetric factors on game results, game simulation experiments were carried out on ER random networks and BA scale-free networks. The experimental results verify the correctness of the game theory and the relationship between the symmetry breaking result and the asymmetry factor. The theoretical model provides theoretical guidance for the study of interaction dynamics between nodes. After that, the relationship between the symmetry breaking between nodes and cascade propagation in the process of load redistribution is studied in this paper. The relationship between game dynamics and cascade dynamics in the network is explored. We construct a cascade model of symmetry breaking game and carry out cascade simulation on BA network and as autonomous network respectively. By adjusting the initial load asymmetry factor and the policy income asymmetry factor in the game, the changes of the maximum connectivity scale and the cascade duration of the network under different attack strategies are observed. Finally, according to the observed time characteristics of cascade evolution, a method is designed to reduce the cascade loss by adjusting the parameters in the game model of symmetry breaking in real time. This result provides a new decision basis and solution for the major accidents caused by cascading faults in a network with behavioral decision ability.
【学位授予单位】：中国科学技术大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：O157.5

【参考文献】