Data-Driven Prediction of Thermal Comfort Using IR Imaging and Wearable Sensor Data

Document Type

Conference Proceeding

Publication Date

1-1-2026

Abstract

As indoor environments increasingly shape people’s daily lives, ensuring optimal thermal comfort and indoor air quality while maintaining energy efficiency remains a significant challenge. This study explores a reinforcement learning (RL)-based approach to optimize HVAC operations in a controlled environmental chamber by integrating physiological and environmental data. The research aims to develop an adaptive control framework that continuously learns from real-time data to enhance occupant comfort and energy efficiency. Experiments were conducted in a controlled environmental chamber where participants engaged in routine tasks while their physiological responses were monitored. Participants included students from various ethnicities and genders. Environmental parameters, including temperature, relative humidity (RH), CO2, CO, PM2.5, and VOC concentrations, were continuously measured. Physiological responses, such as heart rate variability (HRV), electrodermal activity (EDA), and skin temperature, were captured using wearable sensors, while a thermal camera recorded surface temperature variation. An RL framework was developed to optimize HVAC operations based on real-time feedback, adjusting ventilation strategies dynamically to maintain thermal comfort and reduce energy consumption. Data from these sources were analyzed using reinforcement learning algorithms, incorporating a reward-based optimization approach to balance energy efficiency and occupant well-being. The study also investigated the effectiveness of integrating environmental and physiological parameters into the control system, improving its adaptability across different occupancy conditions. The results demonstrated that using RL-based optimization enhances the HVAC performance by dynamically responding to environmental and physiological changes and simultaneously reducing the energy consumption while maintaining occupants’ thermal comfort. Therefore, this approach can be considered as a scalable solution for real-time automated climate control and may provide a pathway toward more sustainable and adaptive indoor environments.

Publication Source (Journal or Book title)

Lecture Notes in Civil Engineering

First Page

530

Last Page

540

This document is currently not available here.

Share

COinS