Implementando adaptação em ambientes inteligentes utilizando sistemas multiagentes e aprendizado por reforço

Abstract

With the advancement of technology, environment control services have become increasingly intelligent and autonomous. To build intelligent systems we need to take into consideration changing situations where the system needs to adapt to the new behaviors observed in the environment. The present work aims to implement a smart home control system using the paradigm of Multiagent Systems and Reinforcement Learning techniques, so that the system can adapt to user preferences and react to changes in user behavior. In this context, this work proposes the use of a Reinforcement Learning technique - specifically the Q-learning Algorithm - as a learning strategy and use of a framework for the development of Multiagent Systems. Therefore, the proposed architecture for the smart home and its functionalities were created. An analysis was performed for the selection of development framework and the system implementation was performed. In addition, this paper presents a case study for the adaptation scenario to an inhabitant of a residence, aiming to evaluate and demonstrate the use of the developed features.