Robot-based measurement of comfort through thermal infrared imaging and wearable sensors

This paper presents an innovative integrated measurement system composed of a social robot, an infrared (IR) sensor and a wearable device (smartwatch) for the assessment of human thermal comfort. The goal of this work is to provide a human-centered methodology that exploits a robotic structure to provide a comfort coaching solution for the end-users in the built environment. For this purpose, physiological parameters such as Heart Rate Variability (HRV) and skin temperature (ts) are measured, in response to different environmental conditions in the office environment. Data were acquired during the summer season, with a dedicated measurement campaign that involved 8 participants. They were exposed to comfort and warm discomfort conditions while the smartwatch and the IR sensor acquired the parameters for 30 minutes. Data analysis was conducted to create suitable input datasets for testing 7 different supervised machine learning (ML) algorithms. The Thermal Sensation Vote (TSV) of the participants was used as the ground truth to evaluate thermal comfort. Results show that in the intrasubject dataset Random Forest (RF) and Naïve Bayes (NB) classifiers can distinguish whether the occupant was in thermal comfort with an accuracy of 93% and 94%, respectively. For the inter-subject comfort evaluation, the average accuracy is 63% for the comfort trial and 49% for the warm discomfort trial. The current research provides a further step in the measurement of thermal comfort, including a robot-based methodology and the use of physiological parameters and ML techniques to interpret human thermal comfort perception in the built environment