Bluetooth Low Energy Based Occupancy Detection for Emergency Management

Abstract

A reliable estimation of an area's occupancy can be beneficial to a large variety of applications, and especially in relation to emergency management. For example, it can help detect areas of priority and assign emergency personnel in an efficient manner. However, occupancy detection can be a major challenge in indoor environments. A recent technology that can prove very useful in that respect is Bluetooth Low Energy (BLE), which is able to provide the location of a user using information from beacons installed in a building. Here, we evaluate BLE as the primary means of occupancy estimation in an indoor environment, using a prototype system composed of BLE beacons, a mobile application and a server. We employ three machine learning approaches (k-nearest neighbours, logistic regression and support vector machines) to determine the presence of occupants inside specific areas of an office space and we evaluate our approach in two independent experimental settings. Our experimental results indicate that combining BLE with machine learning is certainly promising as the basis for occupancy estimation.