Lane departure on combined curves: driver heterogeneity, centrifugal risk, and crash prevention

Wang, Xiaomeng; Zhang, Yujie; Li, Yi; Yi, Xiebowen; Wang, Xuesong; Hao, Guangjie · 2026 · Scientific Reports

DOI: 10.1038/s41598-026-37251-1

archive: archived pipeline: cataloged verified

Get this paper ↗ (DOI — opens at the source; we link to it, we don't host it)

Summary

This study investigates the mechanisms of lane departure crashes on combined curves, addressing a critical road safety challenge given the high fatality rates associated with such incidents. The research aims to quantify how driver heterogeneity and centrifugal force direction influence asymmetric lane departure behaviors, filling gaps in existing literature that often overlooks nonlinear relationships and the specific dynamics of centrifugal force in complex road geometries. The findings are intended to inform the optimization of Advanced Driver Assistance Systems (ADAS), targeted driver training, and visual guidance design. The researchers utilized a high-degree-of-freedom driving simulator to collect vehicle operation data from 36 participants driving on a simulated 24-kilometer mountainous freeway in Hunan Province, China. The test road featured 71 combined curves, categorized into downslope, upslope, sag, and crest types. A total of 948 lane departure events were recorded. Lane departures were classified based on the direction of centrifugal force: In the Direction of Centrifugal Force (IDCF) and Against the Direction of Centrifugal Force (ADCF). The study analyzed average speed, maximum lateral departure, and departure duration distance. To model the nonlinear relationships between driver characteristics and lane departure severity, the authors employed a Multivariate Adaptive Regression Splines (MARS) model. The results revealed distinct patterns in lane departure behavior across different curve types and centrifugal directions. Sag-curves and crest-curves exhibited higher average lane departure frequencies compared to downslope- and upslope-curves. IDCF events demonstrated greater average lateral departure (0.83 m vs. 0.41 m) and longer departure duration distances (70.07 m vs. 58.60 m) than ADCF events, a difference confirmed as statistically significant via Mann-Whitney U testing. The MARS model identified that years of driving, daily driving distance, driving experience, road expert type, and departure duration distance significantly influenced departure severity, with notable interaction effects between driver characteristics and duration distance. Furthermore, speed and road geometry jointly affected departure behavior; for instance, speed significantly impacted lane departure on downslope- and upslope-curves in IDCF scenarios. The study proposed specific safety thresholds: a lane departure duration distance of 35.00–110.00 m and a speed threshold of 86.77–108.63 km/h. These findings provide a quantitative basis for enhancing ADAS systems by incorporating driver-specific characteristics and centrifugal force dynamics, which current systems often lack. The identified thresholds offer concrete parameters for calibrating Lane Departure Warning (LDW) and Lane Keeping Assist (LKA) systems under complex combined curve conditions. Additionally, the results support the development of targeted training programs for high-risk drivers and the optimization of visual guidance designs to mitigate the elevated risks associated with sag- and crest-curves.

Key finding

Asymmetric lane departure on combined curves is dominated by IDCF (with-centrifugal) events, which produce ~2x larger lateral departures and ~20% longer duration distances than ADCF events. Driver experience (years, daily distance, road-expert type) interacts with departure-duration distance to modulate severity, providing thresholds that ADAS lane-departure warning systems and visual-guidance designs can use to calibrate intervention timing on mountainous freeways.

Methodology

simulator

Sample size: N=36 drivers; 948 lane-departure events analyzed.

Provenance

The full processing record for this entry. Every stage of this paper's journey through the pipeline is logged — what ran, with which tool and model, how many attempts it took, and when it last completed. Discovered via canonical_url on 2026-05-03 (8 acquisition events logged).

Summary generated by qwen3.6-27b-prismaquant on 2026-06-10; verification: verified.

Topics

Ranked by relevance to this paper. Hover a topic for its definition.

• lane positioning
• road geometry
• steering pattern
• lane changing
• rail grade crossings
• sex gender

Information type

What kind of knowledge this paper contributes, grouped by family — independent of topic (what it is about) and method (how it was studied).

• Empirical Findings: crash risk outcomes, behavioral performance data
• Theoretical Contribution: computational model