Ant Double Bridge: Pheromone Choice and the Shorter Path

Ants choose between a short and a long bridge by the pheromone earlier ants left. Change the branch ratio, evaporation and feedback and watch the colony settle on the short path.

Switch to 2D

Presets

Playback

Parameters

Camera

Layers

Quality

Model check

The science

The double-bridge experiment. Goss, Aron, Deneubourg and Pasteels (1989) gave Argentine ants two bridge branches between nest and food, one twice as long as the other. After a short time most ants used the short branch, although no ant compares lengths: each ant follows pheromone left by earlier ants. The model here is the choice rule of Deneubourg et al. (1990): P(short) = (k + phis)n / ((k + phis)n + (k + phil)n), with k and n fitted to ant behaviour (k = 20, n = 2).

Why the short branch wins. An ant on the short branch is back and deposits pheromone sooner, so the short branch is refreshed more often and its share of pheromone grows, which draws more ants: a positive feedback with a delay of L / v. Evaporation (d phi / dt = -rho phi) lets the colony forget old trails.

Mean field and checks. For many ants the pheromone obeys the delay equation d phii / dt = q lam [pi(t - taui) + pi(t - 2 taui)] - rho phii (tau = L / v). The page integrates it independently and compares it with the average of 120 stochastic runs; it also checks the choice rule against sampled choices, the path lengths against an analytic circle, the single-branch steady state 2 q lam / rho and exponential evaporation.

Simplifications. An ant chooses once per round trip and uses the same branch back; real ants also choose on the return. Two deposits per round trip, instantaneous deposition at the end of a branch (real ants lay pheromone while walking), constant speed, no ant-to-ant collisions, no random trail losing, no food depletion, and a pheromone that acts only through the branch total. With equal branches the model breaks symmetry at random, as in the original experiment, but the number of ants needed for a stable choice depends on k, n and the flux.

Loading model...
Loading Blender model...
Choice probability against pheromone share
Pheromone on each branch over time
Mean-field short-branch share for several branch ratios
Made with 🟠 Blender · Three.js