Autonomous Mobility on Demand in cities: addressing the minimum fleet problem for an autonomous car sharing service based on massive real trip data

Rapid urbanization has led to escalating traffic congestion, vehicle density, and environmental degradation. Although various policy measures, such as congestion pricing, reduced speed limits, and multimodal incentives, can help, they often prove insufficient to substantially reduce private car ownership. In parallel, fully autonomous “robotaxi” fleets face significant technical and regulatory hurdles, limiting their near term feasibility. This study proposes a valet style Autonomous Mobility on Demand(AMoD) service as a practical alternative. In this scenario, vehicles navigate themselves (empty) to demand locations and are then driven by users for their trips, reducing the complexity associated with fully driverless operations. Using a custom-built simulator and real world travel demand data from over 28,000 private car trips in Milan (Italy), the valet style AMoD fleet size is optimized under the defined quality of service constraints: a maximum waiting time of 20 minutes and fewer than 2% of unmet requests. Results show that a fleet of 2,300 shared autonomous vehicles could serve nearly all trips, achieving more than a 12:1 reduction compared to private car usage. Moreover, the valet style AMoD service is compared with traditional free-floating car sharing, demonstrating significantly higher efficiency and vehicle utilization. These findings highlight the potential of partial autonomy to deliver near term urban transportation services, ensuring congestion relief and environmental benefits, paving the way toward a more sustainable urban people mobility.