Description
Cellular Automata discretise space into a uniform grid of cells and time into ticks; each pedestrian occupies one cell and moves to a neighbour according to local transition rules. Coupled with a floor-field-model to provide a goal-direction bias, CA crowd models scale to thousands of agents at low computational cost while still reproducing macroscopic phenomena (lane formation, jamming transitions, evacuation dynamics).
When it's used
- Large-scale evacuation simulation
- Coupling with floor fields to encode global goals
- Generating synthetic crowd datasets for CSI cross-validation
- Fast-prototyping where Social Force is too expensive
Limitations
- Spatial grid resolution caps physical fidelity
- Discrete velocity set distorts the empirical fundamental diagram
- Sub-second dynamics (jostling, micro-collisions) are smoothed away