To find food, ants follow scent trails, which communicate information via a chemical signal. But it’s an open question how ants find the most efficient routes in the first place, in order to lay down that trail. A swarm of robot ants might have the answer.
A team of scientists programmed sugar-cube sized robots with a few simple rules and placed them in a maze. To simulate the scent trails that ants leave behind, a separate computer was programmed to project a trail of light behind each robot. The light trail faded after a certain amount of time unless a robot passed over it again, which reinforced the trail by making it brighter.
“As the robot stops passing by, the light diminishes,” Simon Garnier, an assistant professor of biology at the New Jersey Institute of Technology, told Discovery News. Garnier worked on the simulation with Maud Combe, Christian Jost and Guy Theraulaz of the Research Center on Animal Cognition in Toulouse, France.
At the start, branches of the maze had no light trail. Every time a robot hit a fork, it would choose a path that deviated the least from the path it was already on. So for the first fork, the robot had to pick one of two paths, each of which would take the robot 60 degrees to the left or right. If the robot “chose” the left path, it was programmed to go to the right at the next fork more often, bringing it back to it’s original “line.” When the robots got to their target, they were programmed to retrace their steps back to the starting point and leave a light trail.
When a light trail was present, the robots were programmed to follow it, and if there were two different light trails, the robots were told to follow the brighter one.
At first, the robots chose a wide range of paths to get to the target. But robots that chose a more efficient route would reinforce the light trail more often. That brighter light trail attracted more robots, which made it brighter still. A light trail that followed a less efficient route would fade.
What surprised the researchers was that the robots managed to navigate the maze using only the pheromone light trail and the programmed rules. They didn’t need to be told what the general shape of the maze was.
This isn’t unlike the common Argentine ants (the ones you usually see in the kitchen). Argentine ants don’t see well and they don’t make calculated decisions about their direction. They just walk forward and choose random paths with the most probable ones – but not the only ones — being straight ahead. Do this dozens, and then hundreds of times while leaving a scent trail to food source and eventually the result is a direct and efficient route.
The fact that the robots oriented themselves in a similar way to real ants suggests that a big brain isn’t necessary to navigate unknown foraging trails, just a few simple rules.
The team published their work in the journal PLOS Computational Biology.
Credit: Simon Garnier / New Jersey Institute of Technology