A model for predicting room occupancy based on motion sensor data
Christos Sardianos, Iraklis Varlamis, Christos Chronis, George Dimitrakopoulos, Yassine Himeur, Abdullah Alsalemi, Fayçal Bensaali, Abbes Amira
Abstract
When designing a large scale IoT ecosystem, it is important to provide economical solutions at all levels, from sensors and actuators to the software used for analytics and orchestration. It is of equal importance to provide non-intrusive solutions that do not violate users' privacy, but above all, it is important to guarantee the accuracy and integrity of the provided solution. In this work, we present a research prototype solution that has been developed as part of an ongoing project called (EM)3. The project involves IoT sensors and actuators, realtime data analytics modules and cutting edge recommendation algorithms in an ecosystem that improves energy efficiency in office buildings. The main concept of the (EM) <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">3</sup> is to recommend energy saving actions at the right moment to the right user. At the core of the (EM) <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">3</sup> vision is to detect when is the right moment for an energy saving action and sensors play a vital role in this. This article focuses on the model that predicts room occupancy using only data from a motion sensor. The predictions of the model, are used to trigger automations and notifications that turn-off office devices (e.g. air conditioning, lights, monitors, etc.) as soon as the office becomes empty, or a few minutes before this happens, in order to further promote efficient energy consumption habits. The evaluation of the model, using data from a camera sensor for validation, demonstrates a very low error rate and a very short delay on the detection of when the room is actually empty.