Description
The decision component of pedestrian-dynamics: which path or exit a walker chooses, given their goal, the geometry, and the observed crowd state. Route choice models complement collision-avoidance models in agent-based crowd-simulation — one selects the global plan, the other handles the local execution. Route choice is the dominant lever in crowd-evacuation outcomes because exit-choice distributions determine queue length more than walking speed does.
Why it's hard
- Routes are chosen on cognitive timescales; the relevant information state is hard to observe.
- Familiarity, signage, herd-following, and density all interact non-linearly.
- Empirical datasets are scarce; lab experiments have limited ecological validity.
- Multi-storey topologies and stairs add discrete decisions on top of continuous fields.
- Adaptive route choice in response to observed congestion is an active research problem.
Common approaches
- Floor-field potentials with cost = travel time + congestion penalty.
- Discrete-choice models (logit-style) over enumerated route options.
- Reinforcement-learning agents for adaptive route choice.
- Empirically calibrated heuristics (shortest-route + nearest-exit + follow-the-crowd).
Source Papers
- duives2013_3924 ↗ — state-of-the-art crowd motion simulation models (route choice covered).
- huang2009_292f ↗ — Hughes' dynamic continuum model (route choice as flow direction).
- feng2021_f56d ↗ — data collection methods for pedestrian behavior.
- lu2026_fb07 ↗ — extended cellular automaton with multi-storey route choice.