Modeling shared autonomous vehicles for large-scale car trip demand: a zonal heuristic approach

Cities with car-dominated travel patterns are exploring Shared Autonomous Vehicles (SAV) as a scalable alternative to private car use. This study evaluates SAV fleet performance in Greater Melbourne using MATSim across four operational scenarios (shared/non-shared, with/without ride rejection). We introduce a Zonal Parallel Insertion Heuristic (ZPIH) to improve shared-ride matching by accounting for spatial demand heterogeneity, and develop a cost-effectiveness index to assess trade-offs across service rate, wait time, occupancy, vehicle kilometers traveled (VKT), and fleet size. Results show that shared operations can reduce fleet requirements by approximately 85–90% relative to the private-car baseline, while unit-capacity scenarios achieve an approximately 81% reduction. Under cost-effective operating conditions, average wait times range from 3–8 min, occupancies exceed 2.5, VKT per ride falls by up to 50%, and vehicle productivity exceeds 40 trips per day. The shared no-rejection scenario also improves spatial equity, demonstrating the potential of demand-aware SAV deployment in car-oriented cities.