Affiliations: [a] School of Electrical and Control Engineering, Shaanxi University of Science & Technology, Xi’an, China | [b] School of Electronic Information and Artificial Intelligence, Shaanxi University of Science & Technology, Xi’an, China | [c] Shaanxi Artificial Intelligence Joint Laboratory, Xi’an, China
Abstract: To solve the problems of sleep disorders such as difficulty in falling asleep and insufficient sleep depth caused by uncomfortable indoor temperature, this paper proposes a deep reinforcement learning method based on deep Q-network (DQN) with human sleep electroencephalogram (EEG) as input to improve human sleep. Firstly, the EEG is subjected to a short-time Fourier transform to construct a time-frequency feature data set, which is used as input to DQN along with temperature. Secondly, the agent performs environmental interaction actions in each time step and returns a reward value. Finally, the optimal strategy for indoor temperature control is formulated by the agent. The simulation results show that this method can dynamically adjust the indoor temperature to the optimal temperature for human sleep, and can alleviate sleep disorders, which has certain practical significance
Keywords: Deep reinforcement learning (DRL), sleep improvement, air conditioning control, deep Q-network (DQN), electroencephalogram (EEG), deep learning
