AINeutralarXiv – CS AI · 18h ago5/10
🧠
DynaOD: Dynamic Origin-Destination Flow Generation with Discrete-to-Continuous Temporal Semantic Modeling
DynaOD is a machine learning framework that generates realistic urban mobility patterns by modeling temporal dynamics through discrete directional trends and continuous evolution, without requiring historical origin-destination data. The approach uses semantic temporal signals to condition pretrained OD generators, achieving better accuracy and distributional fidelity than existing methods with cross-city transferability.