Ant Colony

Stigmergy in action: ants never talk to each other, they only read and write chemical trails in the ground. Pick a tool, then click or drag on the field.

Tools

Brush size 5 Food pile size 7

Parameters

Colony size (max ants) 2500 Evaporation 3.0 %/s Diffusion 15 %

Sensor angle 35° Sensor distance 6

Exploration randomness 20 % Simulation speed 1.0×

Live stats

Food collected / min: 0
Total delivered: 0
Path efficiency: –
Path length estimate: –
Workers / soldiers: 0 / 0
Eggs / larvae / pupae: 0 / 0 / 0
Foraging now: 0
Food left in the field: 0
Predators / ants lost: 0 / 0
Frame time: –

Path length is measured in ant body lengths (1 BL = 3 grid cells) over the last 40 deliveries. Efficiency is the straight-line distance divided by the distance actually walked while carrying food (100 % would be a perfectly straight trail).

What am I seeing? Stigmergy and ant-colony optimisation

Stigmergy is indirect coordination through the environment. An ant changes the world (it deposits pheromone) and that change guides the next ant. No ant has a map, a plan or a leader.
Two chemical fields are simulated on a grid. Ants that are searching lay a to-home trail, so that anyone who later finds food can follow the gradient back. Ants carrying food lay a to-food trail that recruits others. Each deposit is strongest near its source and fades as the ant travels, which creates a gradient to climb.
Each ant samples the field with three antennae (left, centre, right) at the chosen sensor angle and distance and turns towards the strongest signal. Without a signal it performs a correlated random walk; the randomness slider sets how erratic that is.
Evaporation lets the colony forget outdated information (a depleted food pile, a blocked path) while diffusion widens a trail so that ants slightly off the line can still sense it. Too little evaporation and the colony gets stuck on poor routes; too much and trails never build up.
Why the shortest path wins: in the double-bridge experiment (Deneubourg et al., 1990) ants on the short branch complete round trips sooner, so pheromone accumulates there faster. More ants then choose it, which reinforces it further. This positive feedback amplifies a small early difference into a collective decision.
Ant colony optimisation (ACO), introduced by Marco Dorigo in 1992, uses virtual ants and the same ideas (probabilistic choice weighted by pheromone, deposit inversely proportional to tour length, evaporation) to solve routing and scheduling problems such as the travelling salesman problem.
Try it: let a trail form, then paint a wall across it. The ants lose the trail, wander, and re-discover a detour. Rain washes the pheromone away, so the colony has to start again.
A simplified model. The queen turns food stock into eggs (4 s) that become larvae (8 s), then pupae (4 s) and finally workers or soldiers. Soldiers do not forage; they guard the nest and fight spiders. Rates are tuned for readability, not biological accuracy.