Heatmap-Based Decision Support for Repositioning in Ride-Sharing Systems.

The article focuses on the development and evaluation of repositioning heatmaps (RHs) as an intuitive, nonmonetary tool to guide crowdsourced ride-sharing drivers in decentralized repositioning decisions. RHs highlight driver-specific earning opportunities by integrating expected future demand, current and anticipated fleet distribution, and driver location, aiming to reduce service cancellations and improve service availability. An adaptive learning algorithm iteratively updates the expected net demand used in RHs by simulating scenarios and adjusting based on observed cancellations, ensuring the heatmaps reflect dynamic system interactions. Computational experiments using New York City data demonstrate that RHs outperform benchmarks—including nearest repositioning and centralized model predictive control—by lowering cancellations, balancing driver earnings, and maintaining robustness under varying driver compliance levels. The study also discusses the implications of driver heterogeneity and suggests that platform-controlled assignment strategies can incentivize compliance, enhancing system performance and fairness.