AI and the World's Hard Problems · Theoretical computer science · arXiv · 2026-10-23
LLM-Guided Transportation Hub Capacity Planning with Textual Business Inputs
Traditional hub capacity planning models optimize well on quantitative inputs but struggle to incorporate qualitative business context. The proposed framework has a large language model agent iteratively propose capacity decisions from natural-language business descriptions, using a chain-of-thought decision table, and checks each decision through feedback from an optimization model. On a real-world 13-hub freight network in the southeastern US, the approach achieved a 2.8% optimality gap against a hidden ground truth, compared with 11.0% for the traditional optimization model without textual inputs.