| Abstract: |
| This talk explores the use of real-world traffic data for simulating traffic flow in urban areas. The objective is to improve the accuracy and efficiency of traffic simulations by training machine learning models to learn from real-world traffic data. The proposed approaches include learning data-driven models and calibrating existing physics-driven models. Specifically, this talk will introduce some of our latest work on learning to simulate with real-world data, i.e., when the traffic data is sparse and hard to obtain, and the follow-up simulation model and control model facing this real-world data. The findings have implications for traffic control management in real-world settings. |
|