The Future Is Here
We may earn a commission from links on this page

Engineers Build The First Robot Ant Society

We may earn a commission from links on this page.

By using ten tiny robots equipped with light sensors and an exceedingly simple set of rules, scientists have successfully replicated the navigational behavior of colony ants. The breakthrough could result in more efficient transportation systems for humans — but it also gives us a new understanding of how these fascinating insects' hive minds work.


Ants are able to get around their complex transportation networks by using any number of techniques (depending on the species). They can use visual cues, like the forest canopy, the position of the sun, or environmental landmarks along their path. They can also use proprioceptive information — like measuring steps and the number of body rotations. Ants can also follow pheromone trails and exploit social information, such as taking a quick peek at the food loads of their comrades.


But their choice of direction can also depend on the geometrical configuration of connecting points, or nodes. The precise angle of these bifurcations can affect the route followed by an ant in a network of galleries. As a result, these angles can influence the ant’s ability to properly track a pheromone trail to a food source.

Remarkably, the ants always seem to get it right; they do a remarkable job coordinating themselves in even the most awkward and unpredictable pathways — and they persistently choose the most fortuitous path. This has left biologists to conclude that something serious is going on in their decision making and they way they assess their surroundings.

Go ask Alice

To see if this is the case, a research team led by Simon Garnier at New Jersey Institute of Technology’s Swarm Lab, created an experiment in which ant-like robots were confronted with a similar challenge. But instead of dropping pheromone trails, the sugar cube size robots, called Alices, were able to leave light trails that could be detected by light sensors.


The ten robotic ants were then placed in a maze with no pre-existing light trails. They were tested in two types of mazes: one with symmetrical bifurcations and one with asymmetrical ones (which more closely approximated that of a real ant network). Their task was to establish a route between a starting area and a target area in a network of corridors.


The bots were programmed with minimal “exploratory behavior” similar to how real ants operate, namely a random pattern of moving, but in the same general direction.

“The robots were programmed to move in a straight line for a randomly chosen time before turning with a random angle chosen between +30 and -30 degrees,” Garnier told io9. “This provided some flexibility to the behavior of the robot.”


While observing the experiment, the researchers paid attention to the interaction between the displacement of the robotic ants, their trail laying and following behavior, and how it all related to the physical structure of the environment.

To boldly go where other bots have gone before

Needless to say, the robots at first chose the path that deviated the least from their trajectory at each fork in the road. But if the robots detected a light trail — a sign of previous robotic activity — they would turn to follow that path.


“The robots prefer to go where other robots have been before,” said Garnier.


But unlike their previous work with ants, it turned out that the angles of the bifurcations had little to do with their efficiency. They performed the task with great proficiency, and without having to be programmed to identify and compute the physical layout of the bifurcations. All they needed was the pheromone trail of light and their programmed directional random walk. Combined, these simple "traits" directed them to the more direct route between their starting area and the target destination.

“At the beginning of the experiment, when there is no pheromone, ants/robots started using each path equally,” Garnier told us. “But eventually, ants/robots will tend to use shorter paths more and more because it is marked with pheromone more often, and hence will further increase the quantity of pheromone on this path.” It’s a positive feedback loop, says Garnier, as more pheromone leads to more traffic that leads to more pheromone and so on.


But Garnier says this is only half of the story. He wrote to us in an email:

These ants are almost blind, and when navigating their networks they have no idea what is the general direction to their nest. It's like if you were trapped in a heavy blizzard in the countryside of England — so heavy that you cannot read the signs along the road or figure out where is north. And now you have to find your way back to London. The only information you have are the tracks left by other cars. You can follow them like the ants/robots do with the pheromone, but each time you reach a fork you have no idea which path, left or right, will get you closer to London or send you back toward the countryside, or worse will have you trapped into an endless loop (following and reinforcing your own tracks). Ants have found a way around this problem. Their networks of trails are not symmetrical: at a fork, coming back from a food source, the path that makes the smaller angle with their incoming direction is more likely to go toward the nest than the path that makes a larger angle. Also, because this path deviates less from their incoming direction, they are more likely to take it because it requires less effort (smaller rotation).


And in fact, this makes a lot of sense. Argentine ants, the species this study was modeled after, have poor eyesight and move too quickly to make complex decisions about where to go next. As the researchers discovered, it’s actually a very simple process.

Garnier told us that he wasn’t really surprised by the result.

“We had two possible hypotheses: either the ants measure explicitly the geometry of the bifurcation, in which case they would need to use relatively complex cognitive abilities, or they don't and their movement is mostly decided by the physical structure of the environment and their ability to follow the pheromone trail,” he said. “Our experiments show that the second hypothesis is more likely to be true because our robots were not capable of complex cognitive processes and their behavior was remarkably similar to the behavior of the ants.”


Indeed, the discovery will help researchers understand how pheromone trails and the configuration of physical environments affect the behaviors of individual and collective activities in social insects.

You can read the entire study at PLOS.

Images: The Swarm Lab, AntFarmU.