Comparison-Based Ordinal Learning for Proactive Driving Risk Assessment

2026-07-13

AI Summary

This work addresses the limitations of existing driving risk assessment methods, which rely on proxy objectives and struggle to accurately capture true collision risk and relative hazard levels in interactive scenarios, while also lacking frame-level risk annotations. To overcome these challenges, the study introduces ordinal learning into driving risk assessment for the first time, proposing a comparison-based ordinal risk learning framework. This framework generates pairwise supervision signals through temporal evolution, event contrast, and physically grounded counterfactual perturbations, enabling direct modeling of relative risk ordering across frames without requiring numerical risk labels. The approach accommodates multiple parameterization strategies for risk scoring functions and demonstrates significant improvements over state-of-the-art proxy-based baselines on the 100-Car and SHRP2 datasets, achieving notable gains in high-recall risk discrimination, early warning precision, and lead time.

Abstract

Real-time driving risk assessment provides an essential basis for proactive safety by identifying and quantifying the danger of ongoing road interactions before adverse outcomes occur. However, due to the scarcity of collision data and frame-level risk labels, existing driving risk assessment methods often rely on surrogate objectives, which may imperfectly align with true collision risk and not faithfully reflect the relative danger of driving interaction. This paper proposes a comparison-based ordinal risk learning framework that learns collision-relevant risk scores from pairwise supervision in driving data, directly modeling relative risk ordering without requiring numerical frame-level risk labels. We derive pairwise comparisons from three sources of event-structured driving data for such ordinal risk learning: temporal progression within safety-critical sequences, event-level contrast between dangerous and normal interactions, and physics-based counterfactual perturbations. On this basis, instantiations with three risk-scoring function parameterizations are implemented, including directly learning risk scores from comparison data, and aligning existing single or multiple surrogate-based risk models. The proposed framework is evaluated on the 100-Car and SHRP2 naturalistic driving datasets using a proactive collision warning task. Results show that the proposed framework improves high-recall risk discrimination, warning precision, and warning lead time over representative surrogate-based baselines across both in-distribution and out-of-distribution evaluations. These results suggest that the proposed framework can contribute to proactive safety research by providing more reliable risk assessment for automated driving systems and safety-critical driving interactions.

Problem

• driving risk assessment
• collision risk
• ordinal learning
• surrogate objectives
• relative risk ordering

Innovation

• ordinal risk learning
• comparison-based learning
• proactive driving risk assessment
• pairwise supervision
• counterfactual perturbations