Human-like driving behaviour emerges from a risk-based driver model

Kolekar, Sarvesh; Winter, Joost de; Abbink, David A. · 2020 · openalex

DOI: 10.1038/s41467-020-18353-4

archive: archived pipeline: cataloged verified

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

Summary

This paper addresses the lack of a unified, generalizable model for human driving behavior. Existing models are typically scenario-specific (e.g., car-following or curve driving) and fail to capture the underlying principles that allow humans to adapt to diverse driving conditions. The authors propose that human driving is governed by a risk-threshold principle, where drivers maintain perceived risk below an individualized limit rather than optimizing for a specific trajectory. To operationalize this, they introduce the Driver’s Risk Field (DRF), a metric that quantifies perceived risk as the product of the consequence of potential events and the driver’s subjective probability of those events occurring. The DRF is modeled as a two-dimensional field that moves with the vehicle and changes shape based on speed and steering angle, representing the driver’s uncertainty in perception and action. The authors developed a driver model that uses the DRF as a cost function, aiming to achieve a desired speed while keeping the calculated risk below a threshold. To validate the model, the authors first calibrated parameters using human-in-the-loop simulations where a volunteer drove a virtual track in "normal" and "sport" modes. They then tested the model’s predictions against empirical data from literature across seven distinct scenarios: curve radii, lane widths, obstacle avoidance, roadside furniture, car-following, overtaking, and oncoming traffic. The results demonstrate that the DRF model accurately reproduces human-like behaviors across all tested scenarios. In road scenarios, the model correctly predicted curve-cutting behavior, increased lateral deviation on wider lanes, and speed adjustments in response to obstacles and roadside hazards. In traffic scenarios, the model replicated established findings regarding time headway in car-following, braking intensity relative to approach speed, and lateral positioning during overtaking and encounters with oncoming traffic. The model also distinguished between "normal" and "sport" driving styles, with the latter exhibiting higher speeds, lower time headways, and more aggressive maneuvers, consistent with human data. The study concludes that a risk-based cost function, specifically one that accounts for the consequences of perceptual and motor noise, serves as a unifying principle for human driving behavior. This generalizable model offers a scientifically satisfying alternative to fragmented, scenario-specific approaches. The findings have significant implications for the development of automated vehicles, suggesting that incorporating human-like risk assessment mechanisms can improve the naturalness of autonomous driving and potentially enhance user trust in automated systems.

Key finding

A single risk-threshold model based on the Driver's Risk Field reproduces human-like driving behaviour across seven qualitatively different scenarios without scenario-specific switching, supporting the view that satisficing risk minimisation under perceptual/motor noise is a unifying principle of driving behaviour with applications to automated vehicles.

Methodology

simulator

Sample size: Modelling study; for the in-house simulator validation, N=1 (one 25-year-old male driver, 10 normal trials + 10 sporty trials). Other scenario validations rely on previously published datasets cited from the literature.

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 tag_papers on 2026-05-30 (3 acquisition events logged).

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

Topics

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

• risk taking
• mental model of traffic
• situational awareness
• behavioral adaptation risk compensation
• speed choice
• lane positioning

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