Highway global traffic state estimation method based on distributed optical fiber sensing and physical information neural network
By constructing a ResUNet network and optimizing the physical constraint loss function, the problems of noise interference and signal coverage in highway traffic state estimation using distributed fiber optic sensing technology are solved, achieving accurate traffic state estimation and data output to support traffic management.
Patent Information
- Authority / Receiving Office
- CN Β· China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- HEFEI UNIV OF TECH
- Filing Date
- 2026-04-28
- Publication Date
- 2026-07-21
AI Technical Summary
Existing distributed fiber optic sensing technology suffers from noise interference and signal coverage problems in highway traffic state estimation, resulting in unclear vehicle trajectories. Furthermore, physical information neural network models cannot effectively capture multi-scale spatiotemporal dependent features, leading to inaccurate traffic state estimation.
By constructing a ResUNet network, combining distributed fiber optic sensing data, performing gridding and semantic segmentation, optimizing the network using a physical constraint loss function, and extracting multi-scale traffic features, accurate estimation of traffic conditions can be achieved.
It improves the accuracy of traffic state estimation and the generalization ability of the model, outputs reliable data that conforms to the physical laws of traffic flow, and supports traffic incident detection and management.
Smart Images

Figure CN122435786A_ABST