Naslov
Agent-Based Approach for User-Centric Smart Environments

Autori
Mandarić, Katarina ; Skočir, Pavle ; Ježić, Gordan

Izvornik
Smart Innovation, Systems and Technologies (2190-3018) 186 (2020); 37-46

Vrsta, podvrsta i kategorija rada
Radovi u časopisima, članak, znanstveni

Ključne riječi
Internet of Things, Cognitive Internet of Things, Smart Lighting, Artificial Neural Networks, Supervised Learning

Sažetak
Internet of Things (IoT) solutions are becoming irreplaceable in various application domains. IoT enables control over many systems in a smart environment, such as the heating, ventilation and air conditioning (HVAC) system, lighting, and appliances in a smart home or office. By enhancing IoT solutions with a cognitive capability it becomes possible to, for example, adjust ambient conditions according to user preferences without the need for direct user intervention. This functionality constitutes a fore-coming phase in IoT evolution—Cognitive IoT. In this paper, we propose an agent-based smart environment system and compare it to a centralized implementation. In both approaches, feed- forward artificial neural networks are trained under supervision and used to adjust the lighting conditions to the specific user. The agent-based approach offers better preference prediction precision as each user is supported by one agent with a neural network specialized only for his preferences as opposed to the centralized approach where all user preferences are predicted by one neural network. Additionally, the agent-based approach enables easier addition of new users.

Izvorni jezik
Engleski

Znanstvena područja
Elektrotehnika, Računarstvo

POVEZANOST RADA