Effects of heterogeneity in driving manoeuvres on calibrating car-following models at a roundabout

Abstract

Listen

Translate article

This study examines the impact of varied driving manoeuvres – such as entry, circulation, and exit – at roundabouts on the calibration of car-following models. We compare calibrated parameters across three models using car-following trajectories from naturalistic driving data. Subsequently, we assess spatial distributions of calibration errors and evaluate model performances. To ensure fidelity to driving statistics, we validate the distributional similarity with ground truth data through microscopic simulation using Hellinger distance and Kullback-Leibler divergence. Our findings reveal that the car-following models demonstrate improved accuracy when factoring in heterogeneity inherent in roundabout manoeuvres. The Krauss model demonstrates markedly improved accuracy in distributional similarity of gap, velocity, and time-to-collision compared to baseline models. It also reproduces time-to-collision values below one second during circulation manoeuvres. The study underscores the efficacy of the Krauss model, particularly when precisely calibrated for roundabout manoeuvres, in accurately simulating driving behaviours within microscopic simulations.