Indoor Localization Using Ambient FM Radio RSS Fingerprinting: A 9-Month Study

While indoor positioning systems aspire for higher accuracy, their coverage is typically limited to buildings with dedicated hardware. A possible alternative is offered by infrastructure-free positioning methods. In particular, several studies have demonstrated feasibility of indoor positioning using broadcast FM radio signals, which are available in most populated areas worldwide. However, previous work provides little information about long-term performance of FM-based indoor localization.This paper presents a longitudinal study of FM indoor positioning based on received signal strength (RSS) fingerprinting. We evaluate system's performance on a large dataset of real-world FM signals, systematically collected in several large-scale multi-floor testbeds over the course of 9 months. We also investigate the impact of different classifiers, training schedules and fingerprint sizes on localization accuracy. The results demonstrate that well-trained FM-based system can provide reliable indoor positioning even several months after deployment.