A data-driven platform for multimodal positioning and tracking in indoor environments

定位和跟踪人们在室内环境中的活动对于基于位置的服务（LBS）（诸如室内寻路、基于邻近度的广告、劳动力管理、地理围栏、欺诈检测等）是至关重要的。现有的室内定位系统（IPS）在成本、准确性、可用性、对用户与环境之间的交互的理解以及在无设备设置中工作方面存在不足。该项目基于我们在IPS、近距离技术和数据融合方面的丰富经验，开发了一种亚米级精度的多模式IPS解决方案，可用于各种市场（例如零售、健康、仓库等）。我们设想的平台由以下组件组成：（1）分布式摄像头网络，用于监控感兴趣的区域，沿着个人设备上的移动的传感器;（2）实时视觉数据处理和定位，以识别和跟踪对象以及人的活动;以及（3）数据准备和分析引擎，用于清理和管理数据进行分析。在本提案中，我们将解决实时多模态流数据处理所带来的算法和计算挑战，以开发准确可靠的室内定位平台。

Locating and tracking people's activities in indoor environments are crucial to location-based services (LBS) such as indoor wayfinding, proximity-based advertisement, workforce management, geo-fencing, fraud detection, etc. Existing indoor position systems (IPS) fall short in cost, accuracy, usability, understanding of interactions between users and environment, and working in device-free settings. The project builds upon our extensive experience in IPS, proximity technologies, and data fusion to develop a multimodal IPS solution at sub-meter accuracy that can be used in a variety of markets (e.g, retail, health, warehouse, etc.). Our envisioned platform consists of the following components: (1) a distributed camera network that monitors the areas of interest, along with mobile sensors on personal devices; (2) real-time vision data processing and localization to identify and track objects as well as people's activities; and (3) a data preparation and analysis engine that cleans and curates the data for analytics. In this proposal, we will address both algorithmic and computational challenges posed by processing in real-time, multi-modal streaming data to develop an accurate and reliable indoor localization platform.