simulating ant food gathering using pheromone trails
check out the live demo
The ants' logic has no way of knowing where the food or nest are. The only input they have is their antennae, literally two data points in front of them, left and right, that they can sample.
To follow a trail, sample the left and right antennae, and turn in the direction with a stronger signal.
The rules are fairly simple:
- If you're leaving the nest, drop a trail of 'nest trail' pheromone. Try to follow the 'food trail' pheromone.
- If you've just eaten, drop a trail of 'food trail' pheromone. Try to follow the 'home trail' pheromone.
- In either case, if you've lost the scent, wander randomly until you find it again.
If the values are tweaked right, both the foraging ants, and the ants returning home, will converge on a (fairly) optimal trail.
This only works well if several variables are tweaked properly:
- The trails have to fade over time. Otherwise, the map gets saturated with pheromones that are no longer relevant, ie: food that has been eaten up.
- The intensity that the ants lay down trails has to fall off over time. Otherwise, a wandering ant will leave a trail all over the screen, which has the same value as one that happened to walk straight home.
There are plenty of other variables, like ant turning speed, digestion rate, ant count, etc. Most of them are tweakable in the sandbox demo.