To reduce greenhouse gas emissions and temper oil dependence in the transportation sector, solutions such as electric vehicles (EVs) and car sharing have been proposed and promoted. EV sharing, the combination of these technological and operational solutions, aims to overcome hurdles in the way toward mass EV adoption by allowing customers to use EVs without owning them. Several companies, such as Car2Go (a subsidiary of Daimler AG) and Autolib are providing car sharing services with full EV fleets in cities such as San Diego and Paris. A key feature that differentiates these novel service systems from conventional car sharing systems (e.g., Zipcar) is that they allow one-way trips, and in the case of Car2Go, free street parking. With added flexibility, these systems opens up the possibility to meet a larger range of travel needs and show potential to revolutionize urban transportation. In this paper, we study the service region design problem for service provider such as Car2Go. Due to the one-way nature of service, customer adoption of the service critically depends on service coverage, i.e., whether or not their preferred destinations are covered. From the service provider's perspective, the service region has to be designed strategically under uncertainty of customer travel patterns and preferences. We propose a model that supports service region design with fleet sizing decisions to maximize expected profit, based on a distributionally-robust optimization framework to tackle the inherent planning uncertainties. We show that the problem can be transformed into a computationally-tractable mixed-integer second-order cone program (MISOCP). Using this model and real operational data of Car2Go in San Diego, we perform a case study to address various design questions. Our results identify greater profit and higher EV utilization opportunities under properly selected service region and fleet size. It is also shown that EV sharing systems bring more environmental benefits, e.g., savings in carbon emissions, than replacing personal gasoline cars with EV ownership. Furthermore, smaller regional variations in demographics, e.g., population and income levels, lead to more spread service region. Our results also suggest that charging technology advances help to reduce the fleet size and expand the service region while the marginal impacts on service region design is diminishing. This is a joint work with Ho-Yin Mak (HKUST), Ying Rong (SJTU) and Zuo-Jun Max Shen (UC Berkeley).
