Modelling Pedestrian Behaviour in Autonomous Vehicle Encounters Using Naturalistic Dataset

TL;DR

Pedestrians adjust their speed in real-time during autonomous vehicle encounters, a dynamic that is critical for safety in mixed traffic. The paper applies a hybrid ResLogit model to high-resolution, naturalistic NuScenes data, modeling moment-to-moment speed changes using perceptual indicators such as visual looming, collision risk proximity (CRP), and relative speed. Key findings show strong directional asymmetries in CRP effects (frontal versus rear) and nonlinear influences of looming and remaining distance, with relative speed playing a smaller role. These results suggest blended risk-perception and efficiency-driven processes that could inform AV prediction and cueing strategies, though predictive power remains moderate and richer perceptual/contextual features are needed for fuller explanation.

Abstract

Understanding how pedestrians adjust their movement when interacting with autonomous vehicles (AVs) is essential for improving safety in mixed traffic. This study examines micro-level pedestrian behaviour during midblock encounters in the NuScenes dataset using a hybrid discrete choice-machine learning framework based on the Residual Logit (ResLogit) model. The model incorporates temporal, spatial, kinematic, and perceptual indicators. These include relative speed, visual looming, remaining distance, and directional collision risk proximity (CRP) measures. Results suggest that some of these variables may meaningfully influence movement adjustments, although predictive performance remains moderate. Marginal effects and elasticities indicate strong directional asymmetries in risk perception, with frontal and rear CRP showing opposite influences. The remaining distance exhibits a possible mid-crossing threshold. Relative speed cues appear to have a comparatively less effect. These patterns may reflect multiple behavioural tendencies driven by both risk perception and movement efficiency.

Figures (5)

• Figure 1: Directional visual risk surface as a function of collision angle intensity and CTTC, computed as $\text{Intensity}/(1+\text{CTTC})$.
• Figure 2: Spatial choice grid structure in front of each pedestrian position $P_{nt}$.
• Figure 3: Distribution of all observed movement choices across the spatial grid.
• Figure 4: Marginal effects showing how changes in each variable influence the probability of accelerating.
• Figure 5: Elasticities showing the percentage change in the probability of accelerating for a 1% change in each explanatory variable.