Ants are an incredible social insect. In a previous installment, we looked at how a colony can, almost spontaneously, scout out the shortest possible route between their nest and some source of food. This is a problem we humans are very interested in solving, whether to apply it to some Traveling Salesman Problem or to easing traffic flow through telecommunications networks.
In today’s posts, we’re going to take a look at how ants work together to transport objects too large for a single ant to carry. When a group of forager ants land on some huge meal, they must swarm together to lift and carry it back to the nest. In doing so, the colony acts almost as a single organism. In a paper by McCreery et al. (2016), they considered the colony almost as a single organism, largely ignoring individual communications. The ants move in unison, as a group. What is their strategy? How well does it work? Do individuals contribute at all?
To set up this experiment, they assured attraction to some location with cricket, which is apparently french fries for ants. They then replaced it with tuna, which is lighter, once they found it. When the group congregated and lifted the food source, the researchers impeded their path back to the nest. They set a Lego structure (coated in some wonderful product called Insect-a-Slip) down as an obstacle. Then, they observed. How would the group respond when faced with a simple wall? What about a more complex cul-de-sac? What if you trapped them in an enclosed Lego fortress?
When the ants found a wall suddenly dropped in their path, they responded with a mixture of basic strategizing and plenty of good old-fashioned randomness (much like myself in a game of Risk). They tended to approach the wall with bravado, pick a direction, and follow the perimeter. Presumably because of their limited visual capacity, they would randomly decide to switch directions and follow the perimeter of the wall another way. This happened even if the ants were so close to one edge of the wall. Even then, they could randomly decide to try another direction.
It’s doing the best you can with limited information. And sure enough, every ant colony was able to navigate the wall (at least, eventually), and make it back to the nest. How did they do with the cul-de-sac?
Cul-de-sac strategizing was very similar. Find an edge, pick a direction, follow the perimeter, randomly switch directions. This random (ahem, stochastic) behavior did them some good, it seems, in the cul-de-sac. It took them a little longer to navigate out of this and escape to the nest, but they were still able to “solve” the puzzle, tuna in tow, 100% of the time.
Now for the trap. What did the ants do?
The ants seemed to have an eerie sense of premonition here. Although the would sometimes change directions right at the edge of a wall, suggesting limited vision, the congregated ants, when trapped, would very quickly abort mission and scatter, exploring perimeter of their trap frantically. There was some perimeter following inside of it, but mostly without the tuna to weigh them down. In the trap, their priority was escape. Perhaps they figured they could find some safer meals elsewhere.
It was a great experiment with fascinating observations. But again, what’s the point? Why take pleasure in confounding ant colonies who only want to take some tuna home to the nest?
A second paper, by Chen et al. (2015), takes this kind of collective group-effort behavior and attempts to apply it to robots. Their robots of choice were little hockey pucks, somewhat resembling miniature Roombas, and they gave them the task of pushing some object towards a goal.
That object could be a circle, a triangle, or a rectangle. All objects required at least two pushing pucks in order to move. The circles were the easiest to scoot along, but the triangle and rectangle were prone to rotate a little when you pushed one edge, requiring the robots to be a little more dynamic in their response. A pair of pucks pushing a circle could remain in the same place as they moved it along. But a pair pushing a triangle had to constantly adjust to a new location on the perimeter lest the triangle rotate away from their efforts.
Now the question is, how do you tell a swarm of robots to push something? They gave the robotic pucks light sensors, and put different wavelengths on the objects. The goal had red light, and the object had blue light. If a robot was in front of the object, and could see the red light, it better not push the object – it’ll push it away from the goal! So they told those robots to follow the perimeter (as our ants before did). Then, if the robot could find a place where it could no longer detect red light, it would push the blue object. This simple command allowed the robot swarms to successfully push their objects to the goal.
Let’s make things more interesting. What if the goal moves? The researchers set up an obstacle course, and a human behind the scenes moved the goal remotely. The robots were able to navigate this obstacle course, using the same algorithms telling it to push their object whenever they lost sight of the goal.
While they weren’t able to repeat this three-dimensionally, they were able to set up a computer simulation. In the three-dimensional world they set up, swarms of flying robots were able to push floating objects to their goals.
It’s very new technology. But it’s inspired by the world of social insects. Ants may not be very smart, but they’re hard workers, and they know how to get the job done by following a few simple rules. By replicating these principles, we can create advanced new technology. What more might we learn from the social insects of our world, effortlessly congregating into cooperative teams?
Chen, J., Gauci, M., Li, W., Kolling, A., & Groß, R. (2015). Occlusion-Based Cooperative Transport with a Swarm of Miniature Mobile Robots. IEEE Transactions on Robotics, 31(2), 307–321. https://doi.org/10.1109/TRO.2015.2400731
McCreery, H. F., Dix, Z. A., Breed, M. D., & Nagpal, R. (2016). Collective strategy for obstacle navigation during cooperative transport by ants. Journal of Experimental Biology, 219(21), 3366–3375. https://doi.org/10.1242/jeb.143818