The Soria team tested the new approach against a state-of-the-art responsive model on a simulation with five drones and eight obstacles, and confirmed their intuition. In one scenario, the reactive swarms completed their mission in 34.1 seconds, the predictive mission completed in 21.5.
Then came the real demonstration. The Soria team met in small Crazyflie Quadcopters used by researchers. Each was small enough to fit in the palm of your hand and weighed less than a golf ball, but carried an accelerometer, gyroscope, pressure sensor, radio transmitter, and a small motion capture balls, spaced a few inches apart and between the four blades. Readings from the room’s motion capture sensors and camera, which tracked the bullets, were transmitted to a computer running each drone’s model as a ground control station. (Small drones cannot carry the equipment needed to perform predictive control calculations on board.)
Soria placed the drones on the ground in a “start” region near the first tree-shaped obstacles. As she launched the experiment, five drones popped up and quickly moved to random positions in 3D space above the take-off area. Then the helicopters started to move. They slid through the air, between soft green obstacles, above, below and around each other, and towards the finish line where they landed with a slight bounce. No collision. Just an uneventful swarm made possible by a barrage of mathematical calculations updated in real time.
“The results of the NMPC [nonlinear model predictive control] model are quite promising, ”writes Gábor Vásárhelyi, a roboticist at Eötvös Loránd University in Budapest, Hungary, in an email to WIRED. (Vásárhelyi’s team created the responsive model used by Soria, but he was not involved in the work.)
However, notes Vásárhelyi, the study does not address a crucial obstacle to implementing predictive control: the computation requires a mainframe. Outsourcing controls over long distances could leave the entire swarm vulnerable to delays or communication errors. Simpler decentralized control systems may not find the best possible flight path, but “they can work on very small on-board devices (such as mosquitoes, ladybugs or small drones) and scale much, much better with the air. size of the swarm, ”he wrote. Artificial and natural drone swarms cannot have bulky on-board computers.
“It’s a bit of a question of quality or quantity,” continues Vásárhelyi. “However, nature has both.”
“This is where I say ‘Yes, I can,’” says Dan Bliss, systems engineer at Arizona State University. Bliss, who is not involved with the Soria team, is leading a Darpa project to make mobile processing more efficient for drones and consumer tech. Even small drones are expected to get more computational over time. “I take a computer glitch of a few hundred watts and try to put it on a processor that consumes 1 watt,” he says. Bliss adds that creating a swarm of autonomous drones isn’t just about control, it’s also about detection. Embedded tools that map the surrounding world, such as computer vision, require a lot of processing power.
Lately, the Soria team has been working on distributing intelligence among drones to accommodate larger swarms and to manage dynamic obstacles. Predictive-minded drone swarms are, like burrito delivery drones, many years. But it’s not never. Roboticists can see them in their future and, quite possibly, in that of their neighbors as well.
More great WIRED stories