Energy-efficient Airflow Control for Human Thermal Comfort using Object Detection and Monocular Depth Estimation

Abstract

A split-type air conditioner is installed in many households and buildings to produce a comfortable environment. The air conditioner is still mostly operated with occupant's configurations through the remote control. However, it may unexpectedly affect the occupant's health because of the static configurations. Furthermore, improper configurations can reduce the energy-efficient of the air conditioner usage. To address these challenges, we propose a smart control system that is able to automatically control configurations of the air conditioner, including airflow direction, wind speed, and temperature parameters. The smart control system is operated under an embedded box, including a small-embedded device and a camera sensor. It is applied with occupancy detection through object detection and monocular depth estimation through SSD-MobileNetV2 and FastDepth. The occupancy detection integrated environment sensing from the temperature sensor at the living area can help the embedded box determine the optimum control of the air conditioner. The results show that our smart control approach makes a fast comfortable environment and maintains quality of comfort for the entire period of usage with lower energy consumption.