Modelling Uncertainties for Automated and Connected Vehicles in Mixed Traffic

Abstract

Abstract The disruptive nature of automated and connected vehicles (AVs and CAVs) poses increasing risks to infrastructure planning. Predicting their exact impact is impossible because of many unknowns. We address these uncertainties by establishing the upper and lower bounds of performance. An optimisation algorithm was used to guide the simulations so the bounds can be found within a reasonable timeframe. Three AV/CAV models, each given a wider range of parameters than human-driven vehicles (HDVs), were mixed with HDVs in microsimulations. Results show improvements to traffic operations, more so to the freeways than arterial roads, with CAVs offering the most improvement. Our demand sensitivity analysis also estimates the extra demand they can accommodate while maintaining the current delay time.