Ubiquitous and Low-Overhead Floor Identification with Limited Cellular Information

Abstract

Floor identification has gained much attention due to the increasing demand for indoor location-based services, especially prompt emergency response services. Leveraging cellular signals for floor identification has recently been of interest due to the pervasiveness of cellular technology. However, all current systems require information from multiple cell towers concurrently, which is inaccessible in most phones and thus severely limits their deployability. We propose UniCellular , a ubiquitous and deployable floor identification system. UniCellular is the first system to meet the regulatory agencies’ accuracy requirements using received signal strength information from only the serving cell tower. UniCellular relies on a sequence of signal measurements to overcome the limited information available when using only the serving tower. Our system handles multiple challenges affecting accuracy and deployability, including noisy cellular data, overfitting, data collection overhead, and suitability for mobile device deployment. Moreover, UniCellular employs recent advances in deep generative models to improve the system’s robustness to unseen noisy data and reduce data collection overhead. Our extensive experiments verify UniCellular ’s effectiveness in multiple real testbeds, where it correctly estimates the user’s exact floor up to 98.7% of the time using only the serving tower, an improvement up to 310% compared to the state-of-the-art cellular-based system. Furthermore, UniCellular can remain regulatory-compliant with up to 70% reduction in available training data. UniCellular achieves these improvements with over an order of magnitude reduction in the model size compared to the state-of-the-art. This shows UniCellular ’s superior performance while providing a ubiquitous, efficient, and practical solution that meets regulatory requirements.