Robotics

Ant-inspired robot swarm works together to bust out of jail

View 2 Images
Left: Ants inside a corral with sandy walls, which eventually work together to escape. Right: Ant-inspired robots inside their own corral, leaving light field "pheromones" to guide others to move obstacles and escape
Mahadevan Lab/Harvard SEAS
Left: Ants inside a corral with sandy walls, which eventually work together to escape. Right: Ant-inspired robots inside their own corral, leaving light field "pheromones" to guide others to move obstacles and escape
Mahadevan Lab/Harvard SEAS
The researchers programmed the robots to follow simple rules, much like ants do to work together
Mahadevan Lab/Harvard SEAS

One ant is not very smart, but put a bunch of them together and they can achieve some remarkable things. Inspired by this emergent cooperative behavior, Harvard researchers have built robots that can work together to escape from a “prison,” without needing a specific plan.

In a sense, an ant colony runs on algorithms – individual ants don’t specifically think about a task, but they follow a set of instinctual rules that guide them towards complex feats of engineering, like digging tunnel networks or building bridges. They communicate with each other using their antennae and pheromone trails.

The Harvard team started by conducting an experiment with a small group of ants placed inside a corral, a circular trap surrounded by a soft sandy wall, to see how they worked together to escape. At first, the ants all wandered around randomly, but soon some began digging at the walls in scattered spots. Over time, the team found that the ants gave up their solitary digging and would instead group together to work on one tunnel more efficiently, until they eventually managed to break out.

Based on their observations, the researchers built mathematical models of what was happening. As the ants bumped into each other they would communicate via their antennae. Over time, they would start to favor areas where they’d interacted more often, creating a feedback loop where they would then gather in those places more and more. This would lead them to focus their efforts on digging in one spot until they break through, rather than every ant digging their own tunnels.

With this model in place, the team then set out to build robots that followed similar rules. Their robotic ants, which they called RAnts, didn’t give off chemical pheromones but left behind light fields, or “photoromones,” that grew brighter the more times the robots passed by.

The RAnts were programmed to follow three simple rules: they had to follow the gradient of the photoromone field, avoid other robots where the field density was high, and pick up obstacles where the density was high and move them to areas with low density.

And sure enough, these rules allowed the RAnts to cooperate in much the same way as the ant experiment. When placed in their own corral, surrounded by several rings of small obstacles, the robots soon figured out the best escape plan was to work together to focus on one spot.

This technique could be vital for designing swarms of simple robots that can complete complex tasks by working together, and there are plenty of other examples of ant-inspired robots out there. The team says this could potentially be scaled up to dozens or hundreds of robots for a range of use cases, with the added bonus that even if a few robots fail the team as a whole can still get the job done.

“We showed how the cooperative completion of tasks can arise from simple rules and similar such behavioral rules can be applied to solve other complex problems such as construction, search and rescue and defense," said S Ganga Prasath, co-lead author of the study.

The research was published in the journal eLife. The ants and robots can be seen in action in the video below.

Source: Harvard

  • Facebook
  • Twitter
  • Flipboard
  • LinkedIn
0 comments
There are no comments. Be the first!