From GEV to ResLogit: Spatially Correlated Discrete Choice Models for Pedestrian Movement Prediction

TL;DR

This paper compares traditional spatial discrete choice models and a novel neural network-based ResLogit approach for pedestrian movement prediction, finding ResLogit offers superior fit and behavioral coherence in dense spatial decision contexts.

Contribution

Introduces ResLogit, a neural network-enhanced discrete choice model that better captures spatial correlations in pedestrian movement than GEV models, while maintaining interpretability.

Findings

ResLogit significantly outperforms GEV models in fit.

Spatial GEV models offer marginal improvements over multinomial logit.

ResLogit produces behaviorally coherent errors concentrated among neighboring grid cells.

Abstract

High frequency pedestrian motion forecasting when interacting with autonomous vehicles (AVs) can be enhanced through the use of behavioural frameworks, such as discrete choice models, that can explicitly account for correlation among similar movement alternatives. We formulate the pedestrian next step choice as a spatial discrete choice defined by a grid of speed adjustment and heading change. Using naturalistic pedestrian-AV encounters from nuScenes and Argoverse 2 (1 sec decision interval), we estimate a multinomial logit baseline and four spatial generalized extreme value (GEV) specifications (SCL, GSCL, SCNL, and GSCNL). We then compare them to a residual neural network logit (ResLogit) model that learns cross alternative effects while retaining an interpretable linear utility component. Across the evaluated data, spatial GEV structures yield only marginal improvements over multinomial logit, whereas ResLogit achieves a substantially better fit and produces behaviourally coherent errors concentrated among neighbouring grid cells. The results suggest that in dense, high frequency spatial choice sets, learning based residual corrections can capture proximity induced correlation more effectively than analyst specified GEV nesting structures, while maintaining interpretability.