群智感知中动态用户的在线激励机制研究

With the proliferation of mobile devices and the development of wireless communication technology, large-scale crowdsensing systems have drawn people’s great attention in the past years. Incentive mechanisms play a key role in running crowdsensing systems, as they can attract and recruit large amounts of participants. The existing incentive mechanisms mostly assume that the users are synchronized and static, and the system has the prior knowledge of all users. Therefore they cannot be applied in real scenarios where users have spatio-temporal dynamics. To overcome this problem, this project proposes user-first and task-first online incentive mechanisms. The former focuses on the negotiation theory and framework of online mechanisms, and models the system utility based on the utility analysis of users’ locations; while the latter tries to optimize task allocation and improve the system utility based on the online bipartite matching theory. We will theoretically analyze and prove that the proposed mechanisms possess several important theoretical properties, such as strategy-proofness, computational efficiency, and user sovereignty. Finally, we will implement a crowdsensing-based indoor localization system, where the proposed incentive mechanisms will be tested in recruiting dynamic users to automatically update the fingerprint database. The above works build a research framework of incentive mechanisms in crowdsensing, and promotes its application and popularization in the real-world systems.

随着移动终端设备的普及和无线通信技术的发展，大规模群智感知系统被广泛关注。群智感知激励机制能够为群智感知系统提供感知用户基础，是群智感知系统能够正常运行的关键所在。现有激励机制研究多以静态的全局用户信息为前提，无法适应实际应用中用户在时空维度的动态特性。针对该问题，本项目拟研究用户优先和任务优先的在线激励机制，前者将着重研究在线激励机制的交互理论和框架，并基于用户位置的效益分析对感知系统的效能进行建模，后者将侧重感知任务分配与系统效益的提高，基于在线二分图匹配理论和方法进行设计，将任务分配给最合适的用户；拟从理论上分析和证明所提出的激励机制的重要理论性质，包括策略无关特性、计算高效性、用户公平性等；最后对以上理论成果和方法进行实验验证，搭建基于群智感知的室内定位系统，通过激励动态用户实现指纹数据库的自动更新。本项目将完善群智感知激励机制的研究框架，促进其在实际系统中的应用普及。

群智感知激励机制与任务分配是群智感知系统高效运行的关键问题。本项目针对激励机制以及激励机制与任务优化相结合开展研究工作。主要研究内容及意义包括：（1）基于用户位置感知效益分析的激励方法与任务优化。群智感知应用中通常分为两大类型的感知任务，及基于位置的查询任务和基于传感器数据的覆盖任务。本项目研究了融合不同类型感知任务的综合群智感知系统，并通过设计相应的激励和任务优化方法，提高了感知系统的综合效益。所提算法可扩展用于相似的多目标群智感知应用中。（2）基于二分图匹配的激励方法设计。本项目通过引入在线二分图匹配理论和方法，设计了同时具有任务优化分配能力和保证激励机制良好理论性质的激励和任务分配方法。该方法能够推动在线激励与任务分配的进一步融合，提高感知系统在激励和选择用户时的效率，实现感知系统和参与者的双赢。（3）基于系统定价激励的感知系统优化方法。本项目研究了两个基于系统定价激励方式的感知任务分配优化模型：在给定系统激励预算的情况下，最大化每个感知任务的完成概率；在给定感知任务完成概率阈值的情况下，最小化感知系统的激励花费。通过理论分析证明了这两个优化问题的理论性质，并设计了高效的求解算法。所提算法可应用于城市交通路况查询等群智感知应用。