Description
Voronoi tessellation partitions space into cells, one per pedestrian, where each cell is the set of points closest to its agent. The reciprocal of cell area gives a per-agent density estimate that is far less biased than fixed-window kernel estimators, especially in heterogeneous crowds. Voronoi-based density is the standard way to extract a fundamental-diagram from microscopic trajectories and crops up in 3D variants for stairs and multi-level venues.
When it's used
- Local density estimation around each pedestrian
- Empirical fundamental-diagram extraction from BLE trajectories
- Spatial partitioning for cell-level CSI feature aggregation
- Anisotropy analysis in crowd dynamics
Limitations
- Boundary cells become unbounded — need clipping
- Sensitive to localisation noise; small position errors alter cell topology
- 3D extensions are computationally heavier and rarely needed indoors