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

点击翻译按钮获取中文摘要

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.