Mobility-on-Demand Meets Shuttles on the Same Mile

Abstract

Problem definition. We study a point-to-point transportation problem, where the transportation planner can use both shuttles and mobility-on-demand vehicles (MoDs) to serve the incoming random demand of passengers.

A natural example is the problem of employee commute, a setting where employers often use conventional shuttles to transport their employees to work but sometimes also reimburse ride-sharing trips instead.

We model this situation as a dual sourcing problem, where the transportation planner wants to minimize costs, given a maximum expected waiting time of passengers. Shuttles have a high capacity, arrive at a predefined frequency, and have a low cost-per-seat, and MoDs can only transport one passenger, can be called dynamically, and are more expensive.

We characterize the optimal transportation mode to choose, including hybrid strategies. Specifically, we want to understand how and when MoDs can strengthen traditional static transportation.

Methodology/results. We show that there is always a range of target expected waiting times, for which a combination of shuttles and MoDs in a hybrid strategy is optimal. Further, we show that the hybrid strategy is most useful for intermediate (non-limiting) values of shuttle cost-per-seat, shuttle capacity (at fixed shuttle cost) and passenger arrival rates. We also study and simulate alternative models to better understand the role of MoDs in the hybrid strategy and the impact of variability.

Managerial implications. We suggest a novel way to combine static and dynamic transportation inspired by a popularly studied policy in dual sourcing in the supply chain literature. Most interestingly, we show that while hybrid policies are promising, they deviate from traditional wisdom: to decrease the passenger waiting time, the transportation planner should use \emph{fewer} MoDs.

Keywords: mobility-on-demand, transportation, dual sourcing, stochastic modeling