Required Simulated Population Ratios for Valid Assessment of Shared Autonomous Vehicles’ Impact Using Agent-Based Models

Abstract

The impact of shared autonomous vehicles (SAV) on people’s lives, cities, and transportation has been actively discussed in recent years, as the technology is expected to become widespread in the future. Agent-based simulation analysis continues to develop as a method for exploring SAV operation methods and analyzing their impact. Due to high computational costs, when analyzing Mobility-on-Demand services such as SAV via simulation, usually only a fraction of the total population is used (simulated population ratio). However, the size of this ratio might affect the results of the analysis. In this study we evaluated simulated population ratio bias by performing simulations with different ratios and comparing them. For this analysis, a small local city in Japan was set as the target area, and the total travel demand was generated through travel demand synthesis. We also conducted a stated preference survey, estimated a mode choice model, and used it to estimate demand for different travel modes in the simulation. Results revealed that it is necessary to handle travel demand on a scale close to reality when performing a simulation using SAV with dynamic ride-sharing (DRS). Taking the 100% population ratio case as the best estimate, a 60% simulated population ratio was required to reduce bias to less than 10%. On the other hand, when SAV simulation is done without considering dynamic ride-sharing, the bias could be kept under 10% with a simulated population ratio of about 10%. In addition, we also found that for a given population ratio, the magnitude of bias differed for each evaluation index. Especially, empty vehicle kilometers travelled (VKT) ratio is sensitive to simulated population ratio.

Keywords: Shared autonomous vehicles, Ride-sharing, Simulation, Simulated population ratio, Fleet size, Synthetic travel demand