Swarms of drones look magical as they fly through the air, creating visual masterpieces that can be both artistic and useful to viewers. But swarms don’t just come together easily, and usually drones have to be directed by GPS or motion-capture systems to make sure that they fly together well and don’t bump into each other. For the most part, the technology makes sense, but sometimes the fact that drone swarms have to rely on that tech really limits where and when drone swarms can be used.
At the University of Pennsylvania, scientists may have come up with a solution to this limiting factor. The scientists used a camera that looked towards the ground, had a 160-degree field of vision, and an inertial measurement unit to train a 12 drones to fly in a swarm. The swarm flew together so well, and is supposedly the biggest drone swarm to have flown without relying on GPS or motion-capture systems.
One of the scientists responsible for this feat is Dr. Giuseppe Loianno. Loianno believes that this solution is easy to extend to larger swarms, and is also quite cheap to implement. His team thinks that it would be easy to add more drones to the swarm and have them fly just as well together, without compromising the functionality and performance of the swarm. In a massive step towards autonomy, this trial is the first time that several drones have been flown using autonomous navigation, bringing control, planning and perception together to achieve this. The team didn’t hesitate to share the software and hardware used in their project, because they want this to be accessible—they want anyone to be able to fly a swarm of drones without relying on motion-capture technology or GPS.
The drones in the team’s swarm each tracked its own location, and sent its coordinates to a ground station. The station sends back commands to the swarm, directing the drones to change location as necessary. The drones don’t move themselves, but rely on instructions from the ground station to function, which is why drones can be added or removed without any issue.
Loianno’s team hopes to work on loop closure, flying the swarms near dense obstacles, and robustness. The ultimate goal is to develop a solution where each drone can coordinate with the drones around them while also mapping and responding to new environments. Developing a loop closure module will also help with the drift that happens as a result of visual odometry, and will help with allowing drone swarms to fly anyway.