Swarming robots tend to get overwhelmed by their own programming. According to researchers at Georgia Tech, robots they studied which were equipped with safety algorithms to keep them from colliding often ended up freezing in place.
In an article from the IEEE, the writer explains that robots usually have at least two controllers, primary and secondary. The secondary controller is responsible for safety and can override the primary if needed. The robots the researchers were using, the Khepera III from Swiss company K-Team, were doing this too much.
“When you have too many robots together, they get so focused on not colliding with each other that they eventually just stop moving,” said Magnus Egerstedt, director of the Institute of Robotics and Intelligent Machines at Georgia Tech. “Their safety behaviors take over and the robots freeze. It’s impossible for them to go anywhere because any movement would cause their bubbles to pop.”
In order to solve this problem, Egerstedt and his team set about tweaking the algorithms. A minimally invasive safety controller called a control barrier certificate allows the robots to move near one another without going into lockdown.
The researchers tasked four robots with starting from four corners of a room, meeting and circling in the middle, and then moving outward again.
Instead of having personal bubbles, the robots use safe states and barrier certificates to make their movements more precise. Essentially, each robot’s personal bubble is very small, allowing them to progress through the swarm without being frozen with indecision.
The new method “allows the robots to make the minimum amount of changes to their original behaviors in order to accomplish the task and not smack into each other,” said Egerstedt.
This technique could be useful for larger autonomous systems, too. Egerstedt also commented on the possible application of this type of algorithm in self-driving cars.
“Robots are very conservative — they want to make sure they’re safe,” he said. “You couldn’t pack the interstate with self-driving cars with today’s technology.”
Swarms of robots are likely to be encountered in airspace before they’re practical in the form of self-driving cars on the highway. Egerstedt said that the algorithms could be applied to air traffic control as well, directing airplanes in a space that may also include flying robots, and they have already tested it using quadcopters.