Using naturalistic and driving simulator data to model driver responses to unintentional lane departures

Svärd, Malin; Markkula, Gustav; Aust, Mikael Ljung; Bärgman, Jonas · 2024 · Crossref

DOI: 10.1016/j.trf.2023.11.021

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 addresses the need for accurate computational models of driver recovery steering during unintentional lane departures, a leading cause of road fatalities. Such models are essential for conducting prospective safety benefit assessments of Advanced Driver Assistance Systems (ADAS) like Lane Departure Warning (LDW) through counterfactual simulations. The research specifically investigates the timing and amplitude of primary corrective steering adjustments in drift situations on straight roads, aiming to determine how drivers perceive lane departure risk and scale their steering responses. The researchers analyzed three distinct data sets to ensure general validity: the EyesOnRoad naturalistic data set (34 unintentional departures in Sweden), the SHRP2 naturalistic data set (6 crash/near-crash events in the US), and a Run-off-road simulator study (12 artificially induced events in Sweden). Steering behavior was quantified using relative yaw rate as a proxy for steering wheel angle, with individual steering adjustments extracted using a bell-curve decomposition method. The study examined the timing of the first corrective steering action relative to the driver’s last off-road glance and correlated steering amplitude with various visual lane departure risk metrics, including relative yaw angle, splay angle, critical yaw rate, and inverse time-to-lane-crossing (iTLC). Key findings revealed that visually distracted drivers often initiate corrective steering before looking back at the road, indicating that peripheral visual information is sufficient to trigger a response. Regarding steering amplitude, the study found that a simple model describing amplitude as increasing quadratically with the vehicle’s relative yaw angle predicted the primary corrective adjustment better than models based on more complex risk metrics like splay angle or iTLC. This suggests drivers scale their steering input to immediately reposition the vehicle within the lane. While a more complex threshold model based on vehicle orientation also achieved similar fit, the simplicity of the quadratic yaw angle model was notable. The significance of this work lies in providing a robust, evidence-based model for human steering behavior in critical lane departure scenarios. By demonstrating that drivers rely on simple visual cues and scale their response to the vehicle's orientation, the study offers a practical framework for integrating realistic driver models into simulation tools. This enables more accurate prospective evaluations of ADAS effectiveness, moving beyond retrospective crash data to predict how systems might prevent future incidents.

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.

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.

• steering pattern
• lane positioning
• lane changing
• situational awareness
• naturalistic crash near crash
• perceptual countermeasures

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: behavioral performance data
• Methodological Resource: tool software
• Theoretical Contribution: computational model