Collision causal discovery and real-time prediction of freeway tunnels: A novel dual-task approach
Published in Tunnelling and Underground Space Technology, 2025
Recommended citation: Jin, J., Huang, H., Li, Y., & Dai, J. (2025). Collision causal discovery and real-time prediction of freeway tunnels: A novel dual-task approach. Tunnelling and Underground Space Technology, 155, 106216. https://www.sciencedirect.com/science/article/abs/pii/S0886779824006345
’– Employing a structural agnostic model with observational data to identify real-time traffic state causal precursors influencing collisions. This process involves uncovering causal relationships among these causal precursors to elucidate the mechanisms underlying collision occurrences.
– Constructing collision causal graph data to address the interactive relationships among collision causes. The data utilizes causal relationships between causal precursors derived from causal discovery outcomes.
– Developing a causal directed graph convolutional network model based on the collision causal graph to predict freeway tunnel collision risks in real time. The model improves prediction performance by capturing the causal relationship between features.’
Recommended citation: Jin, J., Huang, H., Li, Y., & Dai, J. (2025). Collision causal discovery and real-time prediction of freeway tunnels: A novel dual-task approach. Tunnelling and Underground Space Technology, 155, 106216.