Autonomous taxi fleet relocation: an agent-based analysis of operational trade-offs

Autonomous taxis (AT), which integrate automated vehicles with on-demand services to provide direct door-to-door transport, are transforming urban mobility systems by lowering operating costs and enabling more controlled fleet management. AT services can efficiently match passengers and increase system throughput (e.g., number of served trips). However, AT operations, including passenger matching and proactive relocation, introduce operational uncertainties that could increase costs and travel times due to additional empty vehicle kilometres travelled. We develop an agent-based model to represent vehicle relocations alongside passenger matching and routing for urban AT services and to investigate the AT relocation trade-off between operating costs and operational performance (e.g., wait times, vehicle utilisation). The model is applied to a case study of The Hague, The Netherlands, using detailed road network data and time-dependent origin–destination flows derived from private car trips. Simulation results show that increasing the AT fleet sizes of a company affects service quality of competing operators, resulting in longer average waiting times and up to four additional minutes of travel times due to added traffic. Moreover, relocation generates empty travel, but it occurs during less congested hours, thereby avoiding adding vehicles to road traffic in peak periods. Overall, AT relocation could improve service levels and transport more passengers due to increased availability. An operator using relocation reduces waiting times by about 11% compared to competitors that do not relocate. Profits can also rise by nearly 16% (more than 2,000 euros during morning hours) because more trips are served while the operating costs remain comparatively low.