Promoting carpooling on car-hailing platforms: Order allocation and motivating subsidy

Abstract

This paper investigates an order allocation problem for an online car-hailing platform, including solo-ride and carpooling orders. Compared to solo rides, carpooling provides convenience, reduces emissions, and lowers traveling costs for passengers. However, drivers are unwilling to fulfill carpooling requests due to e.g., extra waiting and detour time to pick up carpooling passengers, and potential disputes and complaints from passengers. Therefore, carpooling brings operational challenges to car-hailing platforms in motivating drivers to serve the carpooling orders and allocating orders to the assign (drivers receive orders reactively) and inform (drivers claim orders proactively) order-dispatching systems. In promoting carpooling services, platforms are willing to provide subsidies to seize the market. In this regard, our study explores the scenario where a car-hailing platform maximizes service-quality-related platform performance by providing subsidies to drivers and optimizing the carpooling order allocation and the matching radius strategies. By taking Didi Chuxing as an example, we build G/M/1-family queueing models to maximize the platform performance measure. Our analysis derives the structure of optimal carpooling order allocation and the threshold subsidy to balance the drivers’ payoff in the two systems at equilibrium. We conduct numerical experiments and sensitivity analysis to simulate close-to-reality cases and find 90% of the carpooling orders should be sent to the assign system with a matching radius of $3\sim 5\mathrm{km}$. For robustness check, we also discuss the cases where the platform’s profit is the objective and the detour time endogenously depends on the matching radius and the order arrival rate. To ensure Pareto improvement for the platform, the drivers, and the passengers, we also apply the $\varepsilon$-constraint method to find the Pareto-improvement sets and the corresponding strategies.