网构软件多策略信任演化博弈机制研究

Internetware is the further abstract for entity sets that running in the network environment. According to the changes of external environment, Internetware can adjust cooperation among entities dynamically. As a result, the evolution of Internetware involves self-organizing characteristic. In the process of evolution, Internetware emerges game of trust behaviors between entities because of the rational thinking of entities. For solving the problems of quantification of the time utility of historical information, division of trust levels, acquisition of profits, modeling of game process, dynamics analysis of trust evolution and optimization of trust strategies, some effective methods are presented to control the selection of behaviors and interactive strategies on account of profit calculation, strategy selection and strategy optimization. Therefore, this project proposes an effective quantification method of historical data based on forgetting characteristic and a division method of trust levels based on fuzzy granular. Meanwhile, this project establishes a calculation model of the global profit function based on the discriminatory service, analyzes the multi-strategy game of trust behaviors, provides a trust evolution model of Internetware based on Wright-Fisher, and establishes an incentive mechanism of trust evolution in light of game. The studies of this project can effectively ensure entities to select trust strategy in the operational stage and make system converge to target state.

网构软件是对运行于网络环境中服务化分布实体集的进一步抽象，它能够根据外部环境变化，动态调整实体间的按需协同，表现具有自组织特征的演化模式。在演化过程中，实体的理性思维使得系统实体间的信任合作行为呈现博弈状态。为通过有效手段控制实体行为及交互策略选取，需以收益计算，策略选择和策略优化为分析问题的基础，解决实体历史行为信息的时效量化，实体信任等级的划分，收益的获取，博弈过程建模，信任演化动力学分析以及信任选择策略优化等6方面问题。基于此，本项目建立基于差异遗忘特征的历史数据时效量化方法，采用基于模糊粒度的实体信任等级划分，建立差异服务下的全局收益函数计算模型，分析网构软件实体信任行为的多策略博弈，提供基于Wright-Fisher的网构软件的信任演化模型并阐述基于博弈的网构软件的信任演化的激励机制。本项目研究能够有效控制网构软件实体在运行阶段偏向信任合作策略，从而使系统快速收敛到目标状态。

本项目以网构软件系统为研究对象，针对系统实体历史数据，建立网构软件实体的历史信息时效量化方法；利用模糊粒度，对网构软件中实体的信任等级的动态划分，提出基于Rough-Fuzzy的网构软件信任度量模型； 对网构软件异构性等特点，提出面向数据流和控制流的网构软件服务模型，并给出网构软件服务实例的可迁移性准则；从博弈论角度，提出网构软件系统中软件实体协作的贝叶斯博弈模型；结合开放网络的复杂性特点，提出了白噪声环境下的网构软件Wright-Fisher多策略信任演化模型，并建立基于演化博弈的激励机制。上述研究表明本项目提出的网构软件多策略信任演化博弈机制，具有较好的适应性和准确性，并促进网构软件实体进行有效协作，可以有效提高网构软件系统的整体可信度。