Construction and Analysis of a Curb Dataset for Last-Mile Delivery in New York City

Yechun Shu

Co-Presenters: Individual Presentation

College: College of Business and Public Management

Major: BS.MANAGEMNT-SUPPLYCHAINMGT

Faculty Research Mentor: Dan Liu

Abstract:

This project will develop a hybrid optimization-and-learning solver to study urban freight operations under strict policy and infrastructure feasibility constraints. A policy- and infrastructure-coupled two-echelon heterogeneous capacitated vehicle routing model (CPA-2E-HCVRP) will be formulated as a Constrained Markov Decision Process (CMDP), enabling explicit representation of how regulations and limited curb access restrict feasible routing and service actions. The approach will implement a two-level framework that separates discrete structural decisions from sequential routing execution. At the structural level, infrastructure and policy configurations (e.g., micro hub activation and staging regimes) will be selected using classical heuristics and mixed-integer programming for small candidate sets and, where appropriate, quantum-inspired QUBO methods for larger combinatorial spaces. At the execution level, a Linformer-based routing policy with strict feasibility masking will support scalable, city-scale decision making.