Crowd dynamics
Coordinated development of a real-time simulation that models pedestrian navigation, collision response, obstacles, congestion, points of interest, and group behavior with force-based vector dynamics.
P—006 / Simulation + analysis
Particle Simulation
01 / Overview
A real-time agent-based crowd simulation in C and SDL2 modeling pedestrian navigation, collision response, obstacle avoidance, congestion, points of interest, and group behavior using force-based vector dynamics.
- Contributors
- Sanjay Kumar, Robert Stott, Beverly Qin
- Project dates
- April 2025 - May 2025
02 / Tools & technologies
05 tools- C
- SDL2
- Agent-Based Modeling
- Statistical Analysis
- R
03 / Technical highlights
Design / ImplementationReproducible agent behavior
Pairwise pedestrian interactions combine Boids separation, alignment, and cohesion with family-group attraction and probabilistic navigation decisions. Seeded randomness makes emergent crowd behavior reproducible.
Concurrent experiments
An automated experiment runner executes independent simulation processes concurrently and records 27 agent-level behavioral and physical variables for reproducible parameter studies.
Statistical analysis in R
Parameter sweeps, principal component analysis (PCA), outlier filtering, correlation analysis, and trajectory visualization reveal how attraction radii and force constants affect crowd behavior.