面向群智感知的高可靠数据收集与筛选关键技术研究

Crowdsensing is a new paradigm which can utilize various sensors equipped in many mobile devices to collect and analyze the data in real world efficiently, providing a kind of pervasive sensing service. Since crowdsensing systems have many advantages, such as low cost and easy to deploy, numerous practical systems have emerged in recent years, gaining a very good utility. Data collection and selection are the critical technologies of crowdsensing, which will dominate the quality of service of crowdsensing systems. However, the existing studies have not fully considered the complex characteristics of crowdsensing systems (such as mobile users have classified reliabilities, crowdsensing tasks are generally related with each other, and most tasks are not suitable for the binary model, etc.), so that they have failed to design efficient data collection and selection mechanisms. Hence, to deal with the above challenges, this project will study some crucial issues of data collection and selection in crowdsensing, including the market-equilibrium-based task pricing mechanisms, the task allocation and data selection mechanisms with users' classified reliabilities, the incentive mechanisms that can encourage users to submit high quality data, and the cheating detection mechanisms, etc. Based on the studies, we will design a series of efficient and highly reliable data collection and selection mechanisms for crowdsensing systems. Moreover, we will evaluate and verify the performances of the designed mechanisms through extensive simulations on real traces.

群智感知系统利用配备在大量移动终端上的各类传感器，可以便捷地获取物理世界中的数据，并对数据进行分析进而形成群智感知服务。由于群智感知系统具有易部署及低成本等特点，近年来涌现出大量实际应用系统，并取得了良好的效果。数据收集与筛选是群智感知中的核心技术，直接决定了群智感知系统的服务质量。然而，现有相关研究并未在充分考虑群智感知系统特点（如用户存在分类可靠性、任务存在相关性以及绝大多数任务属于非0-1模型等）的基础上，设计出高可靠性的数据收集与筛选机制。因此，本课题将围绕群智感知数据收集与筛选中实现市场均衡的最优任务定价、基于用户分类可靠性的任务分配与数据筛选、用户类型转换以及欺骗检测等几个关键技术展开研究，最终设计出一系列机制和算法以实现高效、可靠的数据收集与筛选，并采用真实数据集进行仿真验证，以验证所设计的群智感知数据收集与筛选机制的综合性能。

群智感知系统利用配备在大量移动终端上的各类传感器，可以方便地获取物理世界中的数据，并通过对数据进行分析形成群智感知服务。由于群智感知系统具有易部署和低成本等特点，近年来涌现出大量的实际应用系统，取得了良好的效果。数据采集与筛选是群智感知中的核心技术，直接决定着所能形成的群智感知服务质量。然而，现有相关研究仍存在诸多不足，未能充分考虑群智感知数据收集的特点，解决群智感知中的任务最优定价、基于用户分类可靠性的收集任务分配和数据筛选以及用户的欺骗检测等问题。为此，本课题将在充分考虑群智感知中存在的用户分类可靠性、任务之间存在关联等特性的基础上，围绕实现群智感知任务最优定价、基于用户的分类可靠性的任务分配与数据筛选以及用户的欺骗检测等几个方面展开研究，设计出一系列能够实现竞争均衡的群智感知定价机制、满足群智感知特点的感知数据真值发现机制、基于用户分类可靠性的任务分配机制、用于激励低质量用户提交高质量数据的用户类型转换机制以及群智感知系统中的恶意用户机制等，并在项目资助下发表高质量论文34篇，申请发明专利12项，超额完成了任务。除此之外，本课题还设计开发了多个实际的群智感知系统，用于收集数据并验证所设计机制的有效性，取得了良好的效果。其中，所设计开发的基于群智感知的路面检测系统获得了挑战杯江苏省一等奖。