One-way carsharing service design under demand uncertainty: A service reliability-based two-stage stochastic program approach

This study proposes a two-stage stochastic program approach to jointly optimize the planning and operational decisions for one-way carsharing systems accounting for demand uncertainty. The two-stage stochastic model is constructed in the way that stage-one determines the strategic deployments, including the number of vehicles, personnel and parking lots at each carsharing station, which is referred to as the regular service. Upon the realization of the random demand, stage-two solves the ad hoc service and optimizes vehicle relocation and personnel assignment at the operational level. The ad hoc service is integrated into the model, which serves the excess demand that cannot be covered by the regular service. For computational efficiency, we introduce the concept of service reliability (SR) to decompose the original two-stage stochastic formulation into separable sub-problems. A SR-based gradient solution procedure is then applied to efficiently solve the problem. The performance of the proposed model and algorithm is demonstrated with numerical experiments on a toy and a realistic network. Sensitivity analysis on demand variations and ad hoc service costs is further conducted to evaluate impacts of these key factors on the system performance. Results show that up to 20.49% improvement in the profit can be achieved by the proposed stochastic approach compared to the solution obtained from the historical average model.