Quantifying Self-Explaining Road Performance in Spiral Tunnels: Semantic-Based Evaluation and Simulation Validation

In spiral tunnels, high cognitive load and elevated operational risks are prevalent because of visual monotony and geometric ambiguity. This study develops a quantitative model of the self-explaining level based on the “self-explaining roads” theory, integrating environmental semantic segmentation and a three-level situational awareness model. The model introduces perceptual and comprehension attribute indicators and is validated through driving simulation experiments involving 27 participants across six spiral tunnel scenarios. The results indicate that the proposed model effectively reflects the self-explaining level of roads, with a correlation coefficient with behavioral indicators ranging from −0.234 to −0.326, indicating smoother driving behavior as the self-explaining level increases. As the curve radius increases, the self-explaining level also increases (e.g., 9.9 at radius [R] = 250 m, 21.0 at R = 500 m, 27.6 at R = 970 m). The performance in right-turn scenarios is better than that in left-turn scenarios (20.0 for right turns and 19.0 for left turns). The simulation adopted a left-side driving configuration, consistent with the design assumption for passenger vehicles in the studied tunnel scenario. In this study, drivers were guided to drive in the right lane. Additionally, entrance and exit areas introduce cognitive fluctuations because of abrupt changes in lighting and structure, highlighting the need for targeted optimization in these critical zones. This research provides a quantitative tool and methodological foundation for evaluating the safety of spiral tunnels, paving the way for future exploration into optimized design strategies and underlying cognitive mechanisms.