Nature likes swarms. Birds, ants, bees, brain cells—even people—form swarms when given the chance for reasons that are still not completely understood. They go from being individuals to one cohesive unit that has an internal structure and order. Not only is it fascinating to watch, understanding the supercomplex behavior may help us reveal secrets of life.
In recent years, scientists have tried to create swarm movement in robots by building identical, cheap bots whose sole purpose is to be part of a swarm. But these swarms differed from natural swarms, say a plague of locusts, because they lacked diversity. In the natural world, there are physical, spatial and functional differences between each swarm member. One locust is bigger than its neighbors and therefore has to adjust its movements in minute ways in relation to the locusts around it to keep the swarm cohesive.
Roboswarm science is now addressing this. Researchers say they've developed robots that can swarm while also acting autonomously, so that they're more like animals than droids.
Scientists at the Ecole Polytechnique de Montréal and the Royal Melbourne Institute of Technology say they've developed a platform, dubbed Buzz, that can control swarms made up of diverse bots. The system, which is described in a new paper, allows scientists to control individual robots and the entire swarm at once. Think about it a bit like if scientists had a master switch to control a brain, as well as fine-tuned commands over individual brain cells.
This would allow programmers to subdivide robots into multiple teams, or swarms, the scientists say. Each individual robot can independently communicate with others near it. It allows each bot to track data about its neighbors and about the swarm of which it's a member. These capabilities allow swarms to be created and disbanded, on the fly, with just a few commands. So far, the researchers have only tested Buzz out in simulations.
They tested the ability of virtual robots, running on Buzz, to self-organize into swarms and to calculate the distance they should keep from their neighbors. The distance variable is an important characteristic of swarms. (In locust swarms, for example, if one gets too close to another, an angry neighbor will chomp on it and kill it.) In the first experiment, they assessed how long it took swarms of 10, 100, and 1,000 virtual robots to organize. Through sending each other messages, the swarms were able to self-assemble in roughly 0.15 seconds. It takes them roughly that long to figure out how far away they should be from each other.
Its real test will come when its tested on real-life robots. The researchers note that it's general enough to work on both aerial and terrestrial robots. The platform could also make it easier for engineers to compile "a library of reusable swarm behaviors," according to the paper. That would allow people to create their own swarms with pre-built building blocks, LEGO style.
"We believe that one of the most important aspects of Buzz is its potential to become an enabler for future research on real-world, complex swarm robotics systems," the authors write in the paper. "Currently, no standardized platform exists that allows researchers to compare, share, and reuse swarm behaviors."
If you're interested in coding up your own robot army, Buzz is open-sourced and can be downloaded on Github. Happy swarming!
Daniela Hernandez is a senior writer at Fusion. She likes science, robots, pugs, and coffee.