Comparing and validating models of driver steering behaviour in collision avoidance and vehicle stabilisation

Markkula, G.; Benderius, O.; Wahde, M. · 2014 · OpenAlex-citations

DOI: 10.1080/00423114.2014.954589

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 lack of validation and comparative analysis in existing driver behavior models, particularly regarding steering control during near-crash situations. While numerous models exist for simulating driver control, few have been validated against realistic human data from unexpected near-collision scenarios, and comparisons between alternative models are rare. The authors aim to fill this gap by fitting and comparing four established driver models and two novel simplified models against a dataset of human truck driving behavior during collision avoidance and subsequent vehicle stabilization on low-friction surfaces. The researchers utilized data from a driving simulator study involving 24 truck drivers (half novice, half experienced) who performed repeated collision avoidance maneuvers on a low-friction road ($\mu = 0.25$). The dataset included scenarios with and without Electronic Stability Control (ESC). The study tested four literature-based models: the MacAdam model (minimizing predicted lateral deviation), the Sharp et al. model (weighted sum of path deviations), the Salvucci & Gray model (psychologically motivated control of sight angles), and the Gordon & Magnuski model (satisficing control within boundary limits). Additionally, two new simplified models were developed: an open-loop model for collision avoidance steering and a simple yaw rate nulling control law for vehicle stabilization. These models were parameter-fitted to the human steering data to assess their ability to reproduce observed behaviors. The results indicated that steering to avoid a collision was best described as an open-loop maneuver with a predetermined duration but an amplitude adapted to the specific situation. Conversely, the subsequent vehicle stabilization phase was largely accounted for by a simple yaw rate nulling control law. The established models, which rely on internal vehicle models and path-following optimization, struggled to fit the human data as effectively as the simpler, cue-based approaches. The study found that the ability of the four literature models to fit the data was determined by their capacity to capture these two distinct phenomena: open-loop avoidance and yaw rate stabilization. The significance of these findings lies in the challenge they pose to the concept of internal vehicle models in non-routine driving situations. The authors argue that in near-crash scenarios, driver behavior is better described as direct responses to salient perceptual cues rather than complex internal simulations of vehicle dynamics. This suggests that for simulation-based safety evaluations, such as testing active support systems, simpler models based on perceptual cues may be more accurate and valuable than traditional optimizing models. The paper also highlights methodological issues in model validation, emphasizing the need for realistic near-crash data rather than predefined test-track maneuvers.

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 2026-06-26; verification: verified.

Topics

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

• steering pattern
• mental model of traffic
• situational awareness
• driver post crash behavior
• automation surprise
• anticipation

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
• Theoretical Contribution: computational model, theory or model