Gradient navigation model for pedestrian dynamics.

Area of Science:

• Mathematical Modeling
• Physics of Complex Systems
• Computational Social Science

Background:

• Existing pedestrian dynamics models, such as the social force model, often rely on force-based approaches.
• These force-based models can lead to oscillations and require complex calibration.
• Pedestrian dynamics simulation is crucial for urban planning and safety analysis.

Purpose of the Study:

• To introduce a novel microscopic ordinary differential equation (ODE)-based model for pedestrian dynamics, termed the gradient navigation model.
• To demonstrate the advantages of this ODE-based approach over traditional force-based models.
• To present a theoretically grounded method for parameter calibration.

Main Methods:

• Developed a gradient navigation model using a superposition of gradients of distance functions to determine velocity vector direction.
• Integrated velocity to obtain pedestrian location, forming a system of three ODEs.
• Introduced a parameter calibration method based on theoretical arguments and empirically validated assumptions.

Main Results:

• The model avoids oscillations and inertia effects present in force-based models.
• Smooth derivatives allow for accurate high-order numerical integration, ensuring existence and uniqueness of solutions.
• Simulations show no pedestrian collisions and reproduce emergent phenomena like obstacle avoidance, lane formation, stop-and-go waves, and bottleneck congestion.
• Density evolution and crowd velocity-density relationships align quantitatively with experimental data and fundamental diagrams.

Conclusions:

• The gradient navigation model offers a more robust and efficient framework for simulating pedestrian dynamics.
• Theoretical parameter calibration simplifies model setup and enhances predictive power.
• The model successfully captures complex emergent crowd behaviors, providing valuable insights for real-world applications.