Route choice modelling for cyclists on dense urban networks

Abstract

This paper presents the results of route choice models for cyclists in the city of Zurich. The data includes approx. 4400 cycling trajectories, including approx. 850 from e-bikes. The network is sourced from OSM and the choice set generation is based on a modified version of the BFSLE algorithm. We present descriptive statistics as well as the model results which specifically point out the difference between regular and e-bikes. We provide results of both a simple MNL and a more complex mixed Logit, both estimated in Value-of-Distance (VoD) space, and both suited to directly derive VoD indicators. The results show anticipated effects for cycling infrastructure, speed limits, traffic signals, gradients and traffic volumes. Numerous socio-demographic interaction effects shed light on the taste heterogeneity.