Determining the minimum urban fleet for a valet style autonomous mobility service using real trip data

Private cars now exceed 1.5 billion worldwide and remain parked for most of the day, burdening cities with congestion, land take, and emissions. Fully driverless robotaxi fleets could reverse this trend; however, large-scale European deployment is constrained by regulation, cost, and technology readiness. This study evaluates an intermediate solution: a valet-style Autonomous Mobility on Demand (AMoD) service in which vehicles independently navigate, at low speed, towards calling users, then are driven manually by customers and afterwards relocate autonomously to subsequent users or charging locations, if needed, otherwise select a parking side. A dynamic simulator integrates greedy nearest vehicle assignment to calls, overnight proportional rebalancing, and battery-aware charging. The model replays all the real private‑car trips recorded in one week from 28,369 telematics-equipped vehicles in Milan. Under quality of service constraints, a maximum user waiting time of 20 minutes and fewer than 2% unmet requests, only 2,300 shared autonomous vehicles are required, a twelve‑fold reduction relative to the current private fleet. Serving the same demand with a conventional free‑floating car sharing system would require 7,750 vehicles, confirming the efficiency gain enabled by partial autonomy. By embedding real trip streams, partial autonomy constraints, and operational heuristics into a reproducible framework, the analysis quantifies how a step between private ownership and fully driverless robotaxis can operate feasibly in an urban context. Results suggest that valet AMoD could enable households to dispense with the second or third car typically used for city travel, free curb space, and lower emissions while preserving the individual driving experience and providing a practical bridge toward fully autonomous mobility.