Multi-occupant dynamic thermal comfort monitoring robot system

Nowadays occupant thermal comfort has been a research focus for people’s increasing indoor time and its potential influence on productivity. Efforts have been paid to develop thermal comfort monitoring systems for building environments. Fixed sensors might be less efficient for non-uniform environment distribution. To solve this problem, mobile robots are introduced for occupant thermal monitoring. However, there are still some challenges for thermal comfort monitoring systems, which include multi-view extraction and fusion of human features, dynamic recognition and real-time estimation for multiple occupants, and multi-occupant localization. To fulfill these research gaps, we propose a novel mobile robot based thermal comfort monitoring system, which collects occupant thermal information from RGB-D and thermal images, locates occupants and estimates their thermal comfort in real time. An autonomous robot is designed to automatically recognize occupants, their appearance, wearing cloth, and body temperature from collected images. A machine learning method is trained to estimate each occupant’s thermal comfort. Moreover, the robot can re-identify the occupant in different views, mark their positions and track their trajectory in a real-time map. We conducted an experiment on 20 occupants in 80 h in an office building, and the results showed that our system can estimate the occupant’s thermal comfort in real time with a high ROC-AUC score of 0.84.