Proxy Reconstruction Pre-training for Ramp Flow Prediction at Highway Interchanges
Researchers have developed a new framework called STDAE to improve traffic flow prediction at highway interchanges, particularly where real-time ramp detectors are scarce. This two-stage system uses a Spatio-Temporal Decoupled Autoencoder for pre-training, reconstructing historical ramp data from mainline traffic information. The learned representations are then integrated with forecasting models like GWNet, demonstrating superior performance over existing methods on real-world datasets. AI
IMPACT Enhances AI's capability in real-world infrastructure management by improving traffic flow prediction accuracy.