Probabilistic occupancy level estimation based on opportunistic passive wi-fi localisation

Location and occupancy are information of major interest for ubiquitous applications such as automated services. In this paper, we describe a novel approach for occupancy estimation based on passive localisation of common Wi-Fi devices (such as smart phones), performed without any modification of the devices. We exploit the fact that devices, on which the Wi-Fi is on, regularly emit messages to communicate with reachable access points (AP). We opportunistically extract Received Signal Strength Indicators (RSSIs) using Wi-Fi sniffers and perform a classic fingerprint-based localisation. The location of devices is then used to infer levels of occupancy at the room level. We first evaluate our passive localisation system to assess the possibility to perform accurate enough passive localisation. We then propose a generic probabilistic approach to deduce occupancy levels of zones based on the location of devices that could be applied with any localisation results. We present a prototype currently deployed in our lab demonstrating the feasibility of our approach. We evaluate the performance of our approach in a four-person office in which occupancy ground truth was acquired. Our system is easily deployed, scalable and preserves anonymity of users. Finally, we discuss concerns that such an approach may raise and also its potential.