Engineers from the Georgia Institute of Technology have developed new algorithms that allow swarms of robots to cooperate more efficiently without colliding.
The team demonstrated the algorithms in an experiment captured in a video, in which four robots are asked to swap positions. Without the new system, the robots collide and get stuck in the middle of the test area. In the second attempt, during which the robots use Georgia Tech’s minimally invasive algorithm, the quartet elegantly circle around a centre point in perfect coordination and continue their journey in the requested direction.
"When you have too many robots together, they get so focused on not colliding with each other that they eventually just stop moving," said Georgia Tech roboticist Magnus Egerstedt. "Their safety behaviours take over and the robots freeze.”
This freezing is due to the so-called safety bubbles that engineers programme into the robots. To stay safe, every robot is tasked with keeping a safe distance from other robots while completing the required task.
The Georgia Tech team has essentially managed to decrease the size of the safety bubble to a level where it doesn’t interfere with the robots’ ability to complete the task but still keeps them safe.
"Our system allows the robots to make the minimum amount of changes to their original behaviours in order to accomplish the task and not smack into each other," said Egerstedt.
After having practiced the task with a swarm of four, the researcher then demonstrated the same task with a group of eight. The robots managed to safely complete the task even if one of them stopped cooperating.
Algorithms such as the one by Georgia Tech will be extremely important once self-driving cars really take off.
Even though in experiments Google’s autonomous vehicles have been mostly safe, problems might arise once there are thousands of them on the roads.
“You couldn't pack the interstate with self-driving cars with today's technology," Egerstedt remarked.
The technology could also revolutionise air traffic control. Currently, it is human operators who manage the flows of planes. In future, the planes could be managing themselves using the smart safety-bubble algorithms.