Uber and other ride-sourcing companies rely on multiple types of data to optimize service delivery, and they often use Internet-based platforms to match passengers and drivers through smartphone data. However, this type of supply–demand matching remains challenging due to the need to balance speed, revenue, and quality. HKUST’s Guodong Lyu and colleagues aim to balance these priorities through an algorithm-based “compromise” solution.

“A standard solution is to find an appropriate way to combine the various objectives into a single ‘merged’ objective”, explain the authors. They approach supply–demand matching with the goal of optimizing multiple requirements to achieve a streamlined compromise solution that is both feasible and as close as possible to a “utopia point,” which cannot directly be reached without the laborious process of optimizing each objective separately.

The authors derive their compromise solution to a real-time supply–demand matching problem via multiobjective optimization, an approach in which careful weighting procedures are used to combine multiple objectives, and an online algorithm. “We develop a ride-sourcing simulator,” they explain, “to implement different matching policies.” Specifically, the authors sample passengers’ order information and drivers’ status and service scores and simulate the market environment to identify feasible key performance indicators.

The authors then explore the benefits of their solution for passengers, drivers, and the ride-sharing platform and compare these with the outcomes of other approaches. They find that their solution “dispatches more orders to drivers with higher service scores, sacrificing only a little on pick-up time.” Consequently, their simulated solution enables higher-value passengers to be served, with increased revenues.

“This positive effect could contribute to building a better brand reputation for the platform,” the authors conclude. “It would be interesting to implement this compromise matching policy in the industry and study the long-term effect.”