The F.A.S.T.-Model

A discrete model of pedestrian motion is presented that is implemented in the Floor field- and Agentbased Simulation Tool (F.A.S.T.) which has already been applicated to a variety of real life scenarios.

💡 Research Summary

The paper introduces the F.A.S.T. (Floor field‑ and Agentbased Simulation Tool) model, a discrete, lattice‑based cellular automaton (CA) framework for simulating pedestrian dynamics. The model builds on earlier CA approaches (e.g., Burstedde et al., 2001; Kirchner & Schadschneider, 2002) but adds a three‑level decision hierarchy that separates exit selection, destination‑cell selection, and actual movement.

1. Exit Choice
At the beginning of each simulation round every agent selects an exit E with probability

p_AE = N·(1 + δ_AE·k_E(A)) / S(A,E)^2

where S(A,E) is the Euclidean distance between the agent’s current position and exit E, squared to reflect the area of a circle around the exit. δ_AE equals 1 if the agent chose the same exit in the previous round, otherwise 0; k_E(A) is a personal persistence parameter; N normalises the probabilities. This formulation captures both distance‑based attraction to nearer exits and a “stickiness” effect that models habitual exit choice.

2. Destination‑Cell Choice
Given a chosen exit, an agent may move up to v_max cells per round (typical v_max = 3–6 cells, corresponding to 1.2–2.4 m s⁻¹). All free cells within the reachable neighbourhood constitute candidate destination cells. For each candidate (x,y) the model computes a combined probability

p_xy = N · p_S · p_D · p_I · p_W · p_P

where the partial factors represent five distinct influences:

• Static floor field (p_S) – pre‑computed distance to the exit using Dijkstra’s algorithm; p_S = exp(‑k_S·S_xy).
• Dynamic floor field (p_D) – a vector field left by moving agents; after each move the field at the origin cell is incremented by the movement vector (x‑a, y‑b). The field decays with probability δ and diffuses with probability α to von Neumann neighbours. The influence is p_D = exp(k_D·