Predicting network congestion by extending betweenness centrality to interacting agents.

Area of Science:

• Network Science
• Urban Mobility
• Computational Physics

Background:

• Urban networks experience complex traffic dynamics, especially during peak hours.
• Predicting network activity and congestion is crucial for urban planning and traffic management.

Purpose of the Study:

• To develop a simple yet effective model for predicting edge-level network activity and traffic congestion in urban networks.
• To incorporate interaction effects and agent behavior for more realistic traffic simulations.

Main Methods:

• Extending betweenness centrality approximation with a repulsive mechanism.
• Iterative agent-based modeling with dynamically evolving fastest paths.
• Simulating network saturation and analyzing congestion emergence and connectivity disruption.

Main Results:

• Model accurately predicts speed distribution and congestion maps for London, validated by percolation analysis.
• Identified critical exponents (τ and γ) showing varying dependence on traffic levels and vehicle density.
• Demonstrated inverse proportionality between spatial correlation decay exponent (γ) and vehicle count.

Conclusions:

• The simulation approach effectively captures qualitative and quantitative properties of network loading and peak-hour congestion.
• The model relies solely on topological and geographical network features for accurate predictions.
• Findings provide valuable insights into urban network behavior and congestion dynamics.