Communicating Is Crowdsourcing: Wi-Fi Indoor Localization with CSI-based Speed Estimation
Numerous indoor localization techniques have been proposed recently to meet the intensive demand for location based service, and Wi-Fi fingerprint-based approaches are the most popular and inexpensive solutions. Among them, one of the main trends is to incorporate the built-in sensors of smartphone and to exploit crowdsourcing potentials. However the noisy built-in sensors and multi-tasking limitation of underline OS often hinder the effectiveness of these schemes. In this work, we propose a passive crowdsourcing CSI-based indoor localization scheme, C2 IL. Our scheme C2 IL only requires the locating-device (e.g., a phone) to have a 802.11n wireless connection, and it does not rely on inertial sensors only existing in some smartphones. C2 IL is built upon our innovative method to accurately estimate the moving distance purely based on 802.11n Channel State Information (CSI). Our extensive evaluations show that the moving distance estimation error of our scheme is within 3% of the actual moving distance regardless of varying speeds and environment. Relying on the accurate moving distance estimation as constraints, we are able to construct a more accurate mapping between RSS fingerprints and location. To address the challenges of collecting fingerprints, a crowdsourcing- based scheme is designed to gradually establish the mapping and populate the fingerprints. In C2 IL, we design a trajectory clustering-based localization algorithm to provide precise real-time indoor localization and tracking. We developed and deployed a practical working system of C2 IL in a large office environment. Extensive evaluation results indicate that our scheme C2 IL provides accurate localization with error 2m at 80% at very complex indoor environment with minimal overhead.