移动群智感知中混合激励的博弈融合与优化研究

How to motivate participants to join in sensing task and provide high-quality and reliable sensing data has become an urgent and important problem in mobile crowdsensing. However, the participant who is the object to be motivated always has the characters of rationality, selfishness, mobility and autonomy, which brings many challenges to the design of incentive mechanism. This project intends to study the game fusion and optimization of the hybrid incentive mechanism in mobile crowdsensing based on game theory. The main contents include: Firstly, we construct a finitely repeated task-sensing game model based on multi-objective decision with incomplete information by analyzing the mapping relationship between task sensing and game model. Secondly, we design reputation based rating protocol by using the hybrid incentive mode, analyze the impact of the protocol on individuals’ behavior decision-making in the game and calculate the incentive consistency condition. Thirdly, we analyze the cooperative evolution stability of the task-sensing based decision-making process, use the replicating dynamic equation to verify the robustness of the social optimal strategy, and develop the optimization and adjust program for task-sensing decision-making process. Finally, we also design the experimental platform of hybrid and optimization of incentive mechanism for mobile crowdsensing, and verify the correctness and effectiveness of the hybrid incentive mechanism in practice. The development of the project will provide a scientific and effective solution for the study of cooperative behavior of selfish participants, and promote the rapid and sustainable development of mobile crowdsensing.

如何激励参与者积极参与感知任务并贡献高质可靠的感知数据是移动群智感知亟需解决的关键问题。然而，作为被激励对象的参与者往往具有理性、自私性、移动性、自主性等特征给激励机制的设计带来了诸多挑战。本项目拟基于博弈论研究移动群智感知中混合激励的博弈融合与优化问题。主要内容包括：首先，分析任务感知与博弈模型之间的映射关系，构建基于多目标决策的不完全信息有限次重复任务感知博弈模型；其次，设计基于混合激励方式的信誉评级协议，分析协议对博弈局中人行为决策的影响，计算激励一致性条件；再次，分析任务感知决策过程的合作演化稳定性，采用复制动态方程验证社会最优策略的鲁棒性，制定任务感知决策过程的优化调整方案；最后，设计面向移动群智感知激励机制混合与优化实验平台，从实践层面检验混合激励机制的正确性和有效性。本项目的开展将为激励自私参与者采取合作行为的研究提出科学有效的解决方案，促进移动群智感知快速可持续性发展。

如何有效激励移动群智感知用户积极参与感知任务并贡献高质可靠的感知数据，是当前移动群智感知亟需解决的关键问题。本课题基于博弈论实现混合激励的博弈融合与优化，规范和引导自私参与者的合作行为，克服单一激励方式的局限性，以满足多样化的系统诉求。课题主要从博弈模型构建、激励方式混合设计、合作演化与系统优化、实验平台搭建四个方面开展研究，主要成果包括：.（1）分析了任务感知过程中参与者与感知平台/发布者，以及参与者之间的竞争与合作关系，构建了不完全信息静态博弈模型，基于重复博弈理论将一系列相互对立的静态博弈模型扩展到多对多的有限次重复博弈模型，分析和量化了各激励子目标之间的内在联系，设计了关联各子目标的决策变量及其权重系数，制定出全局统一的多激励目标函数，分析了博弈双方的预期收益及最优策略，并计算出精炼贝叶斯均衡。.（2）设计了激励方式的混合设计与博弈融合算法，将信誉的更新与参与水平和质量评级相结合，遴选出最优的参与者子集，设定了感知数据的报酬支付级别，分析了信誉的稳态分布以及混合激励方式对博弈局中人个体行为决策的影响，计算出激励一致性条件，验证了混合激励方式对任务感知过程中合作性优化问题的可行性，并计算出满足合作均衡最大值的社会最优合作解。.（3）分析了感知决策过程的合作演化稳定性，构造了协作参与者和自私参与者的策略选取复制动态方程组，计算演化稳定均衡解并对其进行稳定性分析，通过解析获得合作涌现对应的损益比，验证了系统期望的最优合作策略的鲁棒性，设计了最优协作策略选取算法，分析了各策略对应的收益及策略选取随时间演化的状态变化，实现了对最优协作策略选取的分析和预测。.（4）构建了面向移动群智感知的基础实验平台，模拟了感知任务发布、参与、完成全过程，针对任务参与者的参与度与所提供的感知数据质量两大衡量激励机制实用价值的重要指标，对混合激励机制进行了验证和评估。