Towards a crowdsourced radio map for indoor positioning system

Abstract

Fingerprinting has become the most popular approach for infrastructure-free indoor positioning systems. But fingerprinting relies on frequent and exhaustive building surveys to build and maintain a radio map of the indoor environment. A concept being developed recently is the crowdsourced indoor positioning system, where users of an indoor area will collectively contribute sensor data collected by personal smart devices to the construction of the radio map. As a proof of concept, this paper proposes and evaluates a promising approach for building the radio map based on a crowdsourced dataset. Evaluation shows that we can achieve a mean positioning error of less than 3 meters base on the crowdsourced map, slightly worse than conventional path survey approach.