| Vision | Last-mile delivery, the final leg of a package’s journey from a distribution hub to a customer’s door, is widely regarded as the most expensive and inefficient stage of the logistics chain. As consumer demand for fast, reliable delivery grows, companies face mounting pressure to route drivers more effectively through unpredictable urban environments. Our vision is to investigate whether modern computational techniques, reinforcement learning and quantum computing, can produce smarter, more adaptive routing decisions for real-world delivery networks. By grounding this research in actual city road data rather than synthetic benchmarks, we aim to provide a realistic foundation for evaluating how these emerging tools perform against the complexities of urban logistics. |
| Mission | This project aims to build and evaluate a multi-faceted delivery optimization system across three integrated components. First, we construct a geospatial data pipeline that extracts real urban road networks from OpenStreetMap and U.S. Census data across six American cities. Second, we develop a reinforcement learning agent that learns adaptive routing strategies through trial and error, handling multi-modal transportation, traffic disruptions, and cost optimization in real time. Third, we implement and benchmark a hybrid quantum-classical dispatch architecture that cascades between quantum eigensolvers and classical methods based on problem size, contributing to an IEEE QuantumWeek 2026 research paper. Together, these components explore complementary approaches to the same core problem: can intelligent systems learn to route deliveries more effectively than traditional methods, and where do quantum techniques offer practical advantage? |
| Team Members | Colin Kula and Miranda Walker |
| Client | Dr. Raychoudhury |
| Site | N/A |