Pompeu Fabra University | Spain

Artur Anguera Ruiz | Marti Sanchez Juanola

The main objective of our idea is to create a smart office environment using machine learning techniques to increase the efficiency of an IOT sensor network. The goal is the implementation of a IOT network prototype to manage the lighting and the air-conditioning system of an office. We consider the air-conditioning system the combination of the heating, cooling and ventilating system. This network would uses sensors and actuators to modify the intensity of the office lights and the curtains length or the temperature combined with the machine learning techniques to achieve a high efficient system capable to predict the employees lighting working patterns and environment temperature preferences in order to reduce the electrical expenses. We also expect our solution to be a benefit for the employee by creating a more friendly working environment. To develop the employees pattern we expect to build a system able to learn directly from the workers by letting them to modify some parameters of the IOT network as the intensity of the light, the color (cold or warm light) and the position of the curtains, the temperature of the air-conditioning system.... Mainly, the machine learning techniques we will employ are based on Reinforcement Learning methods as Q-Learning or SARSA which can be combined with Neural Networks algorithms e.g. the Deep Q-Networks methods.