TU Delft
March 07, 2025, 11:00, Room GR A1 402 (click here for the map)
The rapid growth of on-demand logistics and shared mobility services has underscored the need for intelligent decision-making to enhance operational efficiency and service reliability. This talk presents data-driven solutions to address key challenges in short-term demand prediction and real-time operations for meal delivery services. First, I introduce a predict-then-cluster framework that integrates distributional demand forecasting with adaptive clustering to support dynamic service management. Second, I present a reinforcement learning-based dual-control strategy for real-time order dispatching and idle courier rebalancing, leveraging demand forecasts to optimize delivery efficiency and ensure fair workload distribution among contracted couriers. Beyond on-demand logistics, these approaches also offer insights into their broader applicability for urban on-demand transportation services.
Rina (Jingyi) Cheng is a PhD candidate (2023�) at the Sustainable Urban Multimodal Mobility (SUM) Lab at TU Delft, co-supervised by Dr. Shadi Sharif Azadeh, Dr. ir. Gon�alo Correia, and Prof. dr. Oded Cats. She holds a BSc in Econometrics and Operations Research (Erasmus University Rotterdam, 2021) and an MSc in Computational Science (University of Amsterdam, 2023). Her PhD, part of the Horizon EU project SUM, focuses on real-time demand and availability predictions, operations with learning-based approaches, and complex-system-inspired solutions for shared micro-mobility services.