重复囚徒博弈在无标度网络中的演化

The Evolution of Repeated Prisoner’s Dilemma Game on Scale-Free Network

在P2P和移动P2P网络中,自主节点都需要在有限资源(带宽,电源等)下进行通信和数据共享。如何提高系统的合作水平而减少背叛(搭便车),是个值得深入研究的问题。重复囚徒博弈在生物学、社会学、经济学、信息学等领域正在被广泛的研究,在个体自私的情况下整体涌现合作行为是人们感兴趣的焦点。本文利用遗传算法研究重复囚徒困境博弈在无标度网络中的演化,揭示网络中节点产生合作的相关机制。网络中的节点记忆以前多次和相邻节点的博弈情况,按照一定的编码方法转换成遗传算法中的基因,本文研究了不同记忆长度对合作水平的影响,特征基因的显现分布和基因使用频率,合作节点的度分布情况等。这些研究结论对于设计一个自组织的具有合作机制的系统提供了理论上的支持。

In P2P network and mobile P2P network, independent individuals need to share resources under limited bandwidths and powers. It is valuable to study how to increase the systems’ cooperation while reduce free-riding. Repeated Prisoner’s Dilemma game is widely studied in the fields of biology, sociology, economics and informatics. People’s interests focus on the cooperation emerged in a system that individuals are selfish. We study the iterated games evolved on scale-free network with genetic algorithm, and reveal the cooperation mechanism of nodes in networks. Nodes can remember the game historical strategies and code them into genes. We show that memory length’s effect to cooperation level, and that some characteristic genes and frequently used genes emerge after some generations. We also study cooperate strategy distribution on different degree-nodes. These results may give theoretic support to the design of a self-organized system which can support cooperation.