Traffic Network State Estimation

We are dedicated to designing transferable and self-adaptive models capable of handling dynamic changes and addressing data insufficiency challenges in short-term traffic prediction. Our research explores both model-based and sample-based transfer strategies to uncover the underlying principles of transferable features, ensuring robust and efficient adaptation across diverse traffic environments. By leveraging these strategies, we aim to enhance the generalizability of traffic prediction models, enabling them to operate effectively in complex, volatile, and data-scarce scenarios. Furthermore, our work contributes to improving the accuracy and reliability of traffic state estimation, ultimately supporting more responsive and intelligent transportation systems.

Key Capabilities:

• Transfer learning
• Domain adaptation
• Self-adaptive model