Simulator-based analysis and artificial intelligence examination of driver behaviors in road tunnels

Road tunnels, a critical component of modern transportation networks, have gained strategic importance, particularly with the increasing population and trade volume. Tunnels provide safe transportation in challenging topographies and climatic conditions, while also offering an alternative to road widening in urban areas by reducing environmental degradation and expropriation costs. These structures enhance traffic flow, thereby improving the efficiency of both urban and intercity transportation and contributing to environmental sustainability. Research shows that although the frequency of accidents in tunnels is lower compared to open roads, the severity of accidents tends to be higher, necessitating more comprehensive safety measures in tunnels. In this study, a simulator-based experiment was conducted to analyze driver behaviors and investigate them using artificial intelligence techniques. The research involved a three-stage data collection process, where 134 participants were tested across 12 scenarios. In the first stage, demographic data were collected; in the second stage, driving metrics were gathered through simulator drives; and in the third stage, driver performance was assessed using a simulator performance scale questionnaire. All data analysis and risk assessment processes were carried out using the InTunn.Ai software developed within this study. Through dimensionality reduction analysis, six key components explaining 91.2% of the total variance were identified. Subsequently, a classification model with a 92.31% accuracy rate was developed using hyperparameter optimization. This model categorizes drivers into risk-free, moderate-risk, and high-risk groups. The findings identified abrupt acceleration and braking, lane-changing frequency, compliance with speed limits, and lateral lane deviations as primary distinguishing risk factors. In conclusion, this study not only contributes to the literature on road traffic safety but also serves as a foundation for future research to analyze driver behaviors more effectively and develop advanced risk classification models.