The Necessity of Modeling Location Uncertainty of Fingerprints for Ubiquitous Positioning

Fingerprinting has become a mainstream method for indoor positioning. With the popularization of smart devices, the construction of indoor positioning databases is no longer limited to fingerprints collected in a single carrier; instead, it includes those collected by various platforms, such as IoT devices, smartphones, and robots. To adapt to this new trend, it is key to answer the question: how do fuse fingerprints collected by using various carriers to generate a database for localization? This paper has three contributions: 1) It reveals the necessity of involving the location uncertainty of localization feature measurements. This topic has yet to be considered in the existing research works because they only have fingerprints collected from a single platform. 2) Considering such location uncertainty, this paper proposes an improved database training and location estimation method. 3) This paper presents an approach combining sparse professional fingerprints and dense consumer fingerprints to create a database that is key to the ubiquitous positioning of smart devices. In field tests, the proposed method improved positioning accuracy by over 35% and brought other benefits. The source code of this research is available at https://github.com/zhenqizhen/Location-Uncertainty-FP.git.