Description
Studying the time evolution of crowd state — how density, velocity, and configuration change in response to geometry, intent, and inter-person interaction. Crowd dynamics is the kinematic and rheological facet of crowd-modeling: stop-and-go waves, lane formation, faster-is-slower, arch-and-clog at exits, jamming transitions. The thesis frames CSI-perturbation observations as a dynamical signal whose statistics ought to map onto continuum quantities (velocity field, density gradient).
Why it's hard
- Emergent collective phenomena (lanes, oscillations) appear at intermediate densities and are hard to capture without sufficient sensor resolution.
- Coupling between mechanical contact and decisional behavior produces regime changes that feel discontinuous in observation.
- Boundary effects (doors, corners, columns) dominate dynamics in confined indoor spaces.
- Wireless sensing recovers density but velocity recovery from CSI is far less mature.
Common approaches
- Social Force model and its variants for microscopic dynamics.
- Continuum PDE models inheriting from traffic flow (LWR-style) extended to 2D.
- Cellular automata for fast simulation of evacuation and route-choice dynamics.
- Empirical fundamental diagram fitting (density-velocity relation) from trajectory datasets.