Edge-based graph neural network for ranking critical road segments in a network.

Area of Science:

• Graph Neural Networks
• Network Science
• Transportation Engineering

Background:

• Transportation networks are vital for daily operations and disaster response.
• Identifying critical road segments is crucial for network resilience and emergency management.
• Existing methods face computational challenges in large-scale network analysis.

Purpose of the Study:

• To develop a rapid and efficient method for identifying critical road segments in transportation networks.
• To address limitations in existing approaches for network component ranking and post-disaster recovery.
• To overcome computational overhead in assessing network performance metrics.

Main Methods:

• Representing transportation networks as graphs (roads as edges, intersections as nodes).
• Deploying a Graph Neural Network (GNN) trained on various network changes and disruption scenarios.
• Ranking the importance of road segments based on estimated criticality.

Main Results:

• The GNN model rapidly estimates the criticality rank of road segments in disrupted networks.
• The approach overcomes computational limitations of traditional performance metric calculations.
• Demonstrated effectiveness on synthetic and real-world transportation networks.

Conclusions:

• The proposed GNN approach enables swift identification of critical road segments for improved network resilience.
• Supports rapid decision-making for infrastructure planning and emergency response.
• Offers a computationally efficient solution for large-scale transportation network analysis.