DynaOD: Dynamic Origin-Destination Flow Generation with Discrete-to-Continuous Temporal Semantic Modeling
Researchers have developed DynaOD, a new framework for generating realistic origin-destination (OD) flow patterns in urban mobility. The system translates semantic temporal signals into coherent OD patterns by modeling discrete directional trends and continuous temporal evolution. This approach allows for lightweight integration with existing OD generators and supports cross-city transferability. Experiments show DynaOD outperforms current baselines in predictive accuracy and distributional fidelity. AI
IMPACT Enhances AI's capability in simulating complex urban mobility dynamics, potentially aiding urban planning and traffic management.