An Adaptive Deep Learning Model for Smart Home Autonomous System

Aiming at the problem of enhancing the suggestion and decision-making ability of home automation system under the condition of less user intervention involved, an adaptive control system based on scene perception in smart home is proposed. Firstly, the scenario information is extracted from graphic data by reinforcement learning method, and then the adaptive decision model based on deep learning is constructed, a simulated system was tested based on its ability to predict the best interval present in day for lights and blinds in different rooms, operations of the home equipment is updated according to the user's behavior. Results shows that the proposed model is able to predict the best 15 minute interval in a day for lights and blinds in different rooms of a home without user intervention and user feedback could improve model prediction performance.