The imbalance between vehicle supply and on-demand customers has been a long-standing challenging for central platforms since the advent of ride-sourcing service. In order to address the platform management policy, this work proposed a multi-stage decision model that operates on pricing, matching and routing through the centralized platform with fully compliant drivers. In particular, different from static pricing or myopic regional surge pricing policy, we adopt the origin-destination pricing method involving the demand-supply pattern of both origin and destination. Further, the customer spatial-temporal uncertainty is considered under the formulation of a stochastic programming problem. Then, an Adaptive Dynamic Programming (ADP) based approach is developed for the multi-stage operations, which leads to a practical and computationally tractable algorithm for the problem. Our algorithm is trained and evaluated in the designed simulator based on NYC yellow taxi data and Manhattan road network. It is shown that the total profit can be greatly enhanced compared with traditional static pricing and several surge pricing strategies. Further analysis reveals the impact of drivers' compliance on the total profit of the central platform, which highlights the need for driver incentive policy.

CICTP