Modeling Pedestrian Priority at Unsignalized Crossings Using Machine Learning: A Review

Abstract

Pedestrian safety at unsignalized crossings is a critical concern in urban transport, where priority negotiation often depends on informal interactions rather than strict regulatory control. Over the past two decades, researchers have employed diverse methods to model pedestrian priority, ranging from gap-acceptance theory and statistical regression to simulation-based approaches. However, these methods have struggled to capture the complexity and variability of pedestrian–vehicle interactions across different contexts. Recent advances in machine learning (ML) offer new possibilities by enabling the integration of large-scale, high-resolution datasets and the identification of latent behavioral patterns with greater predictive accuracy. This article presents a literature review of the conceptual, methodological, and policy-oriented research on modeling pedestrian priority at unsignalized crossings. It synthesizes contributions from traditional traffic engineering approaches and highlights the emergence of ML-based methods, including supervised learning, deep learning, and hybrid models. Contextual determinants such as infrastructure design, pedestrian and driver characteristics, and multimodal interactions are also reviewed to frame the broader applicability of predictive modeling. The review identifies critical gaps in data availability, model interpretability, and policy translation, calling for future work on explainable, context-aware ML frameworks and interdisciplinary collaboration. By systematically mapping the state of the art, this literature review underscores the potential of ML-driven insights to inform evidence-based transport policy, enhance pedestrian safety, and support the development of inclusive and sustainable mobility systems.