Driving Safety and Comfort Enhancement in Urban Underground Interchanges via Driving Simulation and Machine Learning

Abstract

Urban transportation systems, particularly underground interchanges, present significant challenges for sustainable and resilient urban design due to their complex road geometries and dense traffic signage. These challenges are further compounded by the interaction of diverse road users, which heightens the risk of accidents. To enhance both safety and sustainability, this study integrates advanced driving simulation techniques with machine learning models to improve driving safety and comfort in underground interchanges. By utilizing a driving simulator and 3D modeling, real-world conditions were replicated to design key traffic safety features with an emphasis on sustainability and driver well-being. Critical safety parameters, including speed, acceleration, and pedal use, were analyzed alongside comfort metrics such as lateral acceleration and steering torque. The LightGBM machine learning model was used to classify safety and comfort grades with an accuracy of 97.06%. An important ranking identified entrance signage and deceleration zones as having the greatest impact on safety and comfort, while basic road sections were less influential. These findings underscore the importance of considering visual cues, such as markings and wall color, in creating safer and more comfortable underground road systems. This study’s methodology and results offer valuable insights for urban planners and engineers aiming to design transportation systems that are both safe and aligned with sustainable urban mobility objectives.