社会成本最小化移动群智感知激励机制研究

发布时间：2018-06-19 05:41

本文选题：移动群智感知 + 激励机制　；参考：《南京邮电大学》2017年硕士论文

【摘要】：移动群智感知被认为是大数据时代最重要的数据采集方式之一,是当前计算机网络研究领域中的一个研究热点,近年已引起了国内外科研人员的密切关注。移动群智感知的核心思想是利用集体的智慧和力量去完成个体很难完成或者需要长期完成的任务,它是以大规模用户参与为前提的。而激励机制对群智系统提升用户参与积极性、保证交易公平性和提高数据质量等方面具有重要作用。同时,移动群智感知中用户的隐私保护问题和感知过程中数据的质量也值得关注。因此,针对不同应用场景和目标设计的激励机制是移动群智感知系统需要解决的一个基本问题。本文主要考虑使得社会成本,即智能手机用户的总成本最小化的激励机制。本文基于该目标针对不同的应用场景,对系统进行建模,对所研究的具体问题进行形式化,采用博弈论方法和相关技术,提出解决用户选择和支付问题的算法。针对时间窗口依赖任务的移动群智感知应用场景,本文分别设计了两类不同的激励机制。在单时间窗口场景中,本文设计出一种基于动态规划方法的最优化算法MST去选取参与用户,而在多时间窗口场景中,本文设计出一种基于贪心方法的算法MMT,可以在多项式时间内获得近似最优解;针对平台有预算约束的应用场景,本文着重考虑一种应用在最大连续时间覆盖且基于预算约束的移动群智感知模型,基于此,本文分别设计了采用单时间窗口激励机制的预算可行性框架和采用多时间窗口激励机制的预算可行性框架,使得平台期望得到的效用最大化。之后,将预算可行的框架扩展至更一般的场景中,即每个手机用户同时可以报多个时间区间的任务;针对移动群智感知过程中感知数据质量需求和用户差分隐私保护的需要,本文考虑一种基于数据质量且满足差分隐私保护的激励机制,将感知数据质量指标和差分隐私保护方法相结合,并且在用户选取阶段分别设计了线性和对数两类得分函数来满足相应性质。通过严格的理论分析和大量实验模拟,证明本文设计的激励机制都能够满足真实性和个人理性。其中,时间窗口依赖任务的激励机制能够使得社会成本最小化,并且MST是单时间窗口情况下解决SOUS问题的最优算法,MMT在多时间窗口中能够在In||+1范围以内获得最优解的近似值。预算约束下最大连续时间覆盖模型的激励机制同样能够使得社会成本最小化,并且满足预算可行性。基于差分隐私的激励机制能够达到近似的社会成本最小化和近似的差分隐私性质。
[Abstract]:Mobile group intelligence perception is regarded as one of the most important data acquisition methods in the big data era, and it is a research hotspot in the field of computer network research. In recent years, it has attracted the close attention of researchers both at home and abroad. The core idea of mobile group intelligence perception is to use collective wisdom and power to accomplish tasks that are difficult for individuals to complete or need to be completed for a long time. It is based on large-scale user participation. The incentive mechanism plays an important role in improving user participation enthusiasm, ensuring transaction fairness and improving data quality. At the same time, the user privacy protection problem and the data quality in the process of mobile group intelligence perception are also worthy of attention. Therefore, the incentive mechanism for different application scenarios and targets is a basic problem to be solved in mobile swarm intelligence sensing systems. This paper focuses on the incentive mechanism to minimize the social cost, that is, the total cost of smartphone users. Based on this goal, this paper models the system for different application scenarios, formalizes the specific problems studied, and proposes an algorithm to solve the problem of user selection and payment by using game theory method and related technology. In this paper, two different kinds of incentive mechanisms are designed for mobile group intelligence perception application scenarios of time-window dependent tasks. In a single time window scenario, an optimization algorithm based on dynamic programming method, MST, is designed to select the participating users. In this paper, we design an algorithm based on greedy method, which can obtain approximate optimal solution in polynomial time. In this paper, we focus on a mobile group intelligence perception model which is applied to the maximum continuous time coverage and budget constraints. In this paper, the budget feasibility framework with single time window incentive mechanism and the budget feasibility framework with multiple time window incentive mechanism are designed to maximize the expected utility of the platform. Then, the framework of budget feasibility is extended to a more general scenario, that is, each mobile phone user can report tasks in multiple time intervals at the same time, aiming at the needs of perceived data quality and user differential privacy protection in the process of mobile group intelligence perception. In this paper, an incentive mechanism based on data quality and satisfying differential privacy protection is considered, which combines perceptual data quality index with differential privacy protection method. And the linear and logarithmic score functions are designed in the user selection stage to satisfy the corresponding properties. Through strict theoretical analysis and a large number of experimental simulations, it is proved that the incentive mechanism designed in this paper can satisfy the reality and individual rationality. The incentive mechanism of time window dependent task can minimize the social cost, and MST is the optimal algorithm for solving the sos problem in the case of single time window. MMT can obtain the approximate value of the optimal solution in the range of in 1 in multiple time windows. The incentive mechanism of the maximum continuous time coverage model under budget constraints can also minimize the social cost and satisfy the budget feasibility. The incentive mechanism based on differential privacy can achieve approximate social cost minimization and approximate differential privacy property.
【学位授予单位】：南京邮电大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TN929.5

【参考文献】