D$^3$-Subsidy: Online and Sequential Driver Subsidy Decision-Making for Large-Scale Ride-Hailing Market
Researchers have developed D$^3$-Subsidy, a novel diffusion-based framework designed for optimizing driver subsidies in large-scale ride-hailing markets. This system addresses the complex challenge of balancing driver supply and passenger demand by employing a hierarchical approach that ensures responsiveness to market fluctuations while adhering to strict subsidy rate caps and low-latency execution. The framework utilizes a prefix-conditioned diffusion model to sample future trajectories and an inverse module to translate these into city-level control signals, with a Lagrangian-dual-derived mapping for efficient dispatch. Real-world A/B testing has confirmed significant improvements in completed rides and gross merchandise value, alongside enhanced compliance with subsidy caps. AI
IMPACT This research introduces a novel AI framework for optimizing dynamic pricing and subsidies in ride-hailing, potentially improving efficiency and profitability for platforms.