Journal of Scientific Engineering Advances

Abstract

Thermal comfort and indoor air quality (IAQ) in public transit play a critical role in protecting passenger health and enhancing travel satisfaction. While bus cabin’s airflow and passenger movement continuously change in real-world operation, capturing these cabin dynamics and comfort is essential for timely and reli able customer feedback and environment monitoring. Most existing approaches rely on supervised learning, which is well suited for classification and prediction when abundant labeled data are available. However, collecting sufficiently large datasets for dynamic bus environments can be expensive, time-consuming, and difficult to scale. In contrast, reinforcement learning (RL) learns an adaptive decision policy from interaction logic rather than fixed supervision, which may allow effective performance even with smaller datasets. In this study, we investigate whether reinforcement learning is suitable for analyzing thermal pattern and IAQ in dynamic bus cabin situations.