Description
The Social Force Model (Helbing & Molnár, 1995) treats each pedestrian as a particle subject to (a) a self-driving force toward a desired velocity, (b) repulsive forces from other pedestrians, (c) repulsive forces from walls and obstacles, and (d) optional attractive forces. Equations of motion are integrated with small Δt to produce continuous trajectories. It is the canonical agent-level model of pedestrian dynamics and the baseline against which most newer crowd models are compared.
When it's used
- Microscopic crowd simulation for evacuation studies
- Generating synthetic ground truth for
wifi-csi-sensing-driven crowd-counting work - Benchmark in
fundamental-diagramcalibration studies - Building block inside hybrid agent-based pipelines
Limitations
- Reproduces the "faster-is-slower" effect but exaggerates particle-like collision artefacts
- Parameter tuning is dataset-specific
- Cannot capture cognitive / strategic behaviour without extensions