MobIntel: Passive Outdoor Localization via RSSI and Machine Learning

Effective planning requires understanding the movement patterns of pedestrians and vehicles. This can be achieved by passive localization of WiFi-enabled devices using RSSI measurements from WiFi probe requests received at nearby sensors. In this paper, we continue our work on a mobility intelligence system (MobIntel) [1] and study the performance of two broad approaches to RSSI-based passive outdoor localization. The first is adapted from traditional active localization methods, including multilateration and fingerprinting. The second relies on machine learning methods, including machine learning-boosted multilateration and classification. We compare the localization performance of the two approaches and find that the machine learning methods consistently outperform the adapted traditional methods. The results demonstrate machine learning methods as promising tools for RSSI-based passive outdoor localization.