Multi-class Fleet Sizing and Mobility on Demand Service.

This paper addresses multi-class fleet sizing and vehicle assignment problem where we aim to provide Autonomous Mobility-on-Demand (AMoD) service using a fleet of heterogeneous vehicles. We present a chain of transportation with three classes of autonomous vehicles including cars, buggies and scooters. Each class of vehicle can access a subset of the network, such that, there are some links exclusive for that particular class. Our fleet management system then assigns available vehicles to trips based on the travel time for passenger pick-up and drop-off, their queue time and accessibility of the road network by the vehicle. Each assignment may consist of a set of vehicles allocated for one trip that is composed of multiple-legs served by different vehicles. For example, first mile pick-up by a scooter, middle mile on a car and last-mile trip on a buggy. We apply a genetic algorithm for heterogeneous fleet sizing and propose a hierarchical structure for travel time optimal assignment of the multi-class autonomous vehicles to passengers. We validated our approach with a range of heterogeneous fleet sizes constrained on the given budget. Our approach is more time efficient than taking a ride on a single-class autonomous vehicle for middle mile plus walking during the first and the last miles. Hence, we provide the convenience of autonomously covering the entire journey using multi-class vehicles with no additional travel or transit delays compared to single-class.