二维平面网格上移动成本群体合作行为研究

从移动成本、收益期望与空间博弈的角度,探讨多主体系统的博弈策略演化与系统涌现特征之间的关系。利用空间演化博弈理论,构建了基于个体移动机制的拓扑结构时刻演变的空间演化博弈模型,分析了当主体具有不同的移动成本与收益期望时系统演化的稳定策略,通过分析稳定策略深入探讨系统中合作簇涌现的机理。仿真结果表明,提高移动成本能够最有效地促进系统合作率,同时中等水平的个体收益期望会进一步促进高移动成本的影响效果。     

空间博弈中基于信号传递的合作行为演化仿真

为了研究信号传递对合作行为演化的影响,利用Repast仿真软件建立合作行为演化模型,根据信号和策略将个体分为利他者、排外者、谄媚者和背叛者,研究信号对合作水平的影响及机制.仿真结果表明,信号是影响合作水平的重要因素,演化过程中发出相同信号的个体形成团簇,发出不同信号的个体相互制约,并且合作水平随着b值的增大而降低.与无信号博弈相比,有信号条件下合作水平明显提高.所得结果对于研究复杂系统中合作行为演化具有一定的指导意义.     

时间尺度与选择倾向性协同作用下的演化博弈模型

针对合作行为的涌现与维持问题，基于演化博弈理论和网络理论，提出了一种促进合作的演化博弈模型。该模型同时将时间尺度、选择倾向性引入到演化博弈中。在初始化阶段，根据持有策略的时间尺度将个体分为两种类型：一种个体在每个时间步都进行策略更新；另一种个体在每一轮博弈后，以某种概率来决定是否进行策略更新。在策略更新阶段，模型用个体对周围邻居的贡献来表征他的声誉，并假设参与博弈的个体倾向于学习具有较好声誉邻居的策略。仿真实验结果表明，所提出的时间尺度与选择倾向性协同作用下的演化博弈模型中，合作行为能够在群体中维持；惰性个体的存在不利于合作的涌现，但是个体的非理性行为反而能够促进合作。     

共演化博弈中的反馈机制

本文研究共演化动力学在合作演化中的作用.系统的状态由个体的策略决定.系统状态与个体的属性共同演化,形成一种反馈机制.特别是当个体能够根据博弈的结果调整社会关系时,这种调整势必影响未来的博弈.这种反馈机制在适当的时间尺度下,总是能够促进合作.首先,分析了个体策略与群组属性共演化的情形,在基于个体选择的层面上,结合溯祖理论和演化集合论,给出了区域性利他行为涌现的条件.其次,给出了结构群体中策略选择的参数判据,将两策略判据和适应动力学结合起来,发现通过调整与收益矩阵无关的参数可以实现性能控制.最后,研究了不同的角色分配方案在最后通牒博弈中对公平行为演化的影响,发现当个体的先行者优势或所配置的资源依赖于先前分配结果时,公平的分配方案及对公平的要求得以建立.共演化这种反馈机制在用博弈论解决编队控制、资源配置方案设计等实际问题时显示出广泛的应用前景.     

基于SPD的网络演化仿真研究

复杂网络结构演化研究中多探讨"如何形成",而忽视了 "为什么这样形成"的问题.基于合作演化的角度,利用空间囚徒困境理论,对社会网络中的个体进行了分类,并建立网络演化中个体选择的微观动力学机制,建立了社会网络的结构演化模型.使用多主体系统仿真工具Repast进行了仿真.利用度分布、聚集系数、平均最短路径及社会总收益作为演化判据,给出了网络演化的仿真结果.结果表明合作机制下的演化网络展现出明显的小世界特性,说明合作机制可以在一定程度上解释现实网络形成的原因.并且指出对于社会整体来说,即使在合作者较少的情况下,也能够通过社会关系的改善极大的提高社会的总体收益.     

理性背叛的推荐合作激励机制

为了促进协作系统中用户的合作行为,激励机制得到了广泛的使用.然而,现有的激励机制往往存在无条件合作策略占优互惠策略的现象,进而抑制了合作的涌现.为了解决这一问题,本文在推荐激励模型上进一步考虑了用户的理性背叛行为.以演化博弈为框架,研究了理性背叛机制在全局平均学习和当前最优学习两种模式下的策略演化特性.结合实际场景,本文还研究了在非完美推荐下理性背叛机制的鲁棒性问题,并且基于余弦相似度提出了一种策略识别方案.最后,通过大量的数值实验与仿真实验,验证了理性背叛机制的理论特性,也展示了该机制在促进合作方面的有效性能.     

具有偏好关系网络结构稳定的群决策协商控制模型

挖掘群决策偏好关系结构信息,提出决策者个体偏好与群体偏好关系以及网络结构稳定的群决策协商控制模型.根据模型提供个体偏好参考基准,计算决策者及群体的偏好相容性测度.通过决策者建立、取消(断开)或加强与其他决策者的链接以及对偏好信息进行调整的策略建议,促使决策者个体偏好、群决策偏好关系网络结构正向演化,在模型框架下保证群体偏好网络结构稳定,达到群体偏好信息相容性极大的目的,为进一步信息集结提供依据.     

面向联邦学习激励优化的演化博弈模型

针对联邦学习中参与者虚报训练成本导致激励不匹配的现象,提出了面向联邦学习激励优化的演化博弈模型.首先在联邦学习系统中建立了联邦参与者-联邦组织者演化博弈模型,设计模型质量评估算法对参与者提交的模型进行质量评估,去除低质量模型的同时量化参与者训练成本.然后结合信誉度指标提出优化的激励分配方法,通过求解演化博弈的稳定策略得到不同初始状态下的最优收益策略.最后仿真实验表明参与者激励收益方面,与平均分配法和个体收益分享法相比诚实参与者的收益提升了70%和57.4%,虚报参与者收益降低了65%和69.5%,策略选择方面,所提模型能合理选择收益策略.     

惩罚因子在重复囚徒困境博弈中的研究

针对囚徒困境博弈中收益矩阵参数无法动态更新的问题,提出一个带惩罚因子的囚徒困境博弈模型。该模型中的个体可以根据自身的策略,动态修改收益矩阵中的参数,在每轮博弈之后个体根据邻居收益更新自己的策略。仿真结果表明,惩罚因子可以有效地促进合作策略的涌现,另外发现,合作策略的涌现不仅与惩罚因子有关,而且与网络个体的初始策略有关。当社团内部的个体采取相同策略,社团之间采取不同策略时,更有利于合作策略的涌现。最后,惩罚因子还可以提高参与者的平均收益。     

自组织P2P 网络分布式合作激励机制研究现状

自组织P2P网络本质上的节点自治及理性特征决定了网络系统目标与节点自身利益的不一致性,合作激励机制能够引导节点采用利他的行为策略,弥合系统与个体之间的利益冲突．以提高自组织网络可用性为目标,探讨了多种有效的、实用的分布式合作机制,总结了自组织P2P网络合作激励机制的设计中存在的问题和研究方向．     

Strategy evolution driven by switching probabilities in structured multi-agent systems

Evolutionary mechanism driving the commonly seen cooperation among unrelated individuals is puzzling. Related models for evolutionary games on graphs traditionally assume that players imitate their successful neighbours with higher benefits. Notably, an implicit assumption here is that players are always able to acquire the required pay-off information. To relax this restrictive assumption, a contact-based model has been proposed, where switching probabilities between strategies drive the strategy evolution. However, the explicit and quantified relation between a player's switching probability for her strategies and the number of her neighbours remains unknown. This is especially a key point in heterogeneously structured system, where players may differ in the numbers of their neighbours. Focusing on this, here we present an augmented model by introducing an attenuation coefficient and evaluate its influence on the evolution dynamics. Results show that the individual influence on others is negatively correlated with the contact numbers specified by the network topologies. Results further provide the conditions under which the coexisting strategies can be calculated analytically.     

Endeavouring to be in the good books. Awarding DTN network use for acknowledging the reception of bundles

This paper describes an incentive scheme for promoting the cooperation, and, therefore, avoiding selfish behaviours, in Delay Tolerant Networks (DTN) by rewarding participant nodes with cryptographic keys that will be required for sending bundles. DTN are normally sparse, and there are few opportunistic contacts, so forwarding of other’s bundles can be left out. Moreover, it is difficult to determine the responsible nodes in case of bundle loss. The mechanism proposed in this paper contributes to both problems at the same time. On one hand, cryptographic receipts are generated using time-limited Identity Based Cryptography (IBC) keys to keep track of bundle transmissions. On the other hand, these receipts are used to reward altruistic behaviour by providing newer IBC keys. Finally, these nodes need these IBC keys to send their own bundles. When all nodes behave in a cooperative way, this incentive scheme works as a virtuous circle and achieves a Nash equilibrium, improving very much the network performance in terms of latency. The scheme is not difficult to implement, and it can use an already existing IBC infrastructure used for other purposes in a DTN.     

机会路由中激励合作机制的博弈分析

为了解决机会路由中因保存能量、内存或其他资源而拒绝合作的节点自私性问题,基于微观经济学和博弈论,通过构建合理的效用函数,建立一种促进合作的激励机制,使节点无法通过不真实的反馈信息为自己牟利,有效减少理性节点的作弊行为。仿真结果表明,该合作增强机制能够有效引导理性节点诚实合作,从而提高网络吞吐量。     

Wright–Fisher multi-strategy trust evolution model with white noise for Internetware

A trust evolution model plays an important role in ensuring and predicting the behaviors of entities in Internetware system. Most of the current trust evolution models almost adopt expertise or average weight method to calculate entities’ trust incomes, and focus on two strategies (‘full trust’, ‘full distrust’) to analyze trust behaviors. In addition, the researches on dynamics evolution models fail to consider the factor of noise, and cannot effectively prevent free-riding phenomenon. In this paper, a trust measurement based on Quality of Service (QoS) and fuzzy theory by considering timeliness of history data is proposed to improve the accuracy of trust measurement results. Furthermore, a trust evolution model based on Wright–Fisher and the evolutionary game theory is proposed. This model considers multi-strategy and noise problems to improve the accuracy of prediction and adaptability of model in complex networks. Meanwhile, in order to solve the free-riding problem, and improve the trust degree of a system, an incentive mechanism is established based on evolutionary game theory to inspire entities to select trust strategies. The simulation results show that this model has good adaptability and accuracy. In addition, this model can effectively improve network efficiency, and make trust income reach an optimal value, so as to improve trust degree of a system.     

P2P soft security: On evolutionary dynamics of P2P incentive mechanism

This paper thoroughly investigates the evolutionary dynamics of soft security mechanism, namely, reciprocity-based incentive mechanism, in P2P systems based on Evolutionary Game Theory (EGT). By soft security mechanism, it means social control mechanisms to overcome peers’ selfish (rational) behaviors, and encourage cooperation in P2P systems. Specifically, there exist three strategies in P2P systems: always cooperative (ALLC), always defect (ALLD) and reciprocator (R). Instead of existing work which take it for granted that, like ALLC users, R users did not bear any information-seeking cost, we assume small reciprocation cost, and study generalized mutation-selection dynamics. Our contributions are threefold: firstly, we prove and illustrate that, in a well-mixed P2P structure, ALLD is the only strict Nash equilibrium; secondly, we infer the specific condition under which evolution dynamics exhibits rock-scissors-paper oscillation in a structured P2P population. That is, the population cycles from ALLD to R to ALLC and back to ALLD; finally, we theoretically illustrate that the intensity of selection plays an important role in the evolutionary dynamics of P2P incentive mechanism. That is, when the intensity of selection is relatively weak and reciprocation cost limits to zero, the time average can be mostly concentrated on reciprocator. In brief, considering the existence of reciprocation cost and the small mutation in P2P incentive mechanisms, unlike existing work, it is impossible to simply achieve the “absolute cooperative” in P2P incentive mechanisms. On the other hand, stochastic evolution in P2P incentive mechanism with finite population and network structure still favor reciprocation.     

An adaptive group decision making framework: Individual and local world opinion based opinion dynamics

Opinion dynamics (OD) models, which simulate individuals’ opinion evolution process on social network to analyze the final state of opinion distribution in a group, usually differ from each other due to the differences in social network evolution rules and opinion evolution rules. However, most existing social network evolution rules and opinion evolution rules usually cannot characterize the comprehensive influence of key factors such as neighbors and opinion differences in social relationships. To fully consider the properties of social network evolution and improve the efficiency of consensus reaching process in group decision making, this paper introduces the concept of local world opinion derived from individuals’ common friends, and then proposes an individual and local world opinion-based OD model. In the proposed model, social network evolution is jointly determined by the distance between individual opinions and network structure similarity. The pair of individuals with the largest consensus improvement space are then suggested to adjust their opinions by using an adaptive individual opinion adjustment mechanism. Finally, detailed simulation results are provided to demonstrate the convergence of the proposed model and analyze different parameters’ effects on the stabilized time steps and the number of stable state opinion clusters.     

基于激励机制的网络攻防演化博弈模型研究

随着网络信息系统的日益复杂化,网络的安全性和用户隐私性引起了人们的高度重视,寻找能够维护网络安全、分析和预判网络攻防形式的新技术尤为重要.由于演化博弈理论的特性与网络攻防的特性较为契合,因此,本文对网络环境进行了分析,构建网络攻防场景,并在惩罚机制的基础上引入激励机制,提出了基于激励机制的攻防演化博弈模型.通过给出群体不同的问题情境,利用复制动态方程对局中人的策略选取进行演化分析.另外,在第三方监管部门对局中人管理的基础上,分析不同攻击时长时攻击群体的演化规律,证明攻击具有时效性.通过激励机制对防御群体策略选取的影响以及引入防御投资回报,来进一步证明增加激励机制的可行性.根据实验验证表明,本文提出的攻防演化博弈模型在不同的问题情境下均可达到稳定状态并获得最优防御策略,从而有效减少防御方的损失,遏制攻击方的攻击行为.     

基于声誉的移动众包系统的在线激励机制

在大数据环境下,对移动众包系统的研究已经成为移动社会网络（MSN）的研究热点。然而由于网络个体的自私性,容易导致移动众包系统的不可信问题,为了激励个体对可信策略的选取,提出一种基于声誉的移动众包系统的激励机制——RMI。首先,结合演化博弈理论和生物学中的Wright-Fisher模型研究移动众包系统的可信演化趋势;在此基础上,分别针对free-riding问题和false-reporting问题建立相应的声誉更新方法,从而形成一套完整的激励机制,激励感知用户和任务请求者对可信策略的选取;最后通过模拟实验对提出的激励机制的有效性和适应性进行了验证。结果显示,与传统的基于社会规范的声誉更新方法相比,RMI有效地提高了移动众包系统的可信性。     

Incentive mechanism for selfish nodes in wireless sensor networks based on evolutionary game

A Wireless Sensor Network (WSN) is made up of a mass of nodes with the character of self-organizing, multi-hop and limited resources. The normal operation of the network calls for cooperation among the nodes. However, there are some nodes that may choose selfish behavior when considering their limited resources such as energy, storage space and so on. The whole network will be paralyzed and unable to provide the normal service if most of the nodes do not forward data packages and take selfish actions in the network. In this paper, we adopt a dynamic incentive mechanism which suits wireless sensor networks based on the evolutionary game. The mechanism emphasizes the nodes adjust strategies forwardly and passively to maximize the fitness, making the population in the wireless sensor network converge to a cooperative state ultimately and promoting the selfish nodes cooperating with each other such that the network could offer normal service. The theoretical analysis and simulation results show that the proposed model has better feasibility and effectiveness.