Comprehensive Behavior Spectrum Analysis of Risky Driving Behaviors on Mountain Curves

Constrained by complex topography, mountain roads with tight bends and steep slopes show inherently high‐risk traits, with accident rates several times higher than on ordinary roads. Dynamic driving behaviors and complex environmental interferences have long hindered risk assessment due to fragmented data and static analysis. Previous studies rely on single‐variable analysis or laboratory simulations, lacking real‐world on‐road data validation, thus keeping risk assessment at an empirical level and impeding precise high‐risk behavior tracking. To address this, on‐road experiments were conducted on eight typical sharp curves of Chongqing’s X429 County Road, collecting 512 groups of trajectory data from 32 experienced drivers via high‐precision sensors. First, a mountainous driving behavior risk spectrum is constructed. Key findings show sharp steering is the core risk factor (risk weight: 0.3558), outpacing other behaviors. Correlation coefficients among speeding, sudden speed and direction changes, and lane deviation are all < 0.3, demanding independent control measures. Risky driving behaviors aggregate spatially: speeding and speed changes predominantly occur on precurve straights and postcurve slopes, while lane deviation and sudden direction changes occur within curves and at transition points, aligning with accident hotspots. In this study, the first data‐driven risk spectrum for mountainous sharp curves is established, overcoming traditional single‐factor limitations. Findings support road design optimization and intelligent early warnings, fostering proactive risk control in mountain roads.