Theres a good chance youve seen a drone swarm. Maybe not in person, but probably televised during a New Years celebration. A drone swarm occurs when a large number of the flying robots take to the skies in sync.

It isnt a coincidence that they almost always fly in open outdoor areas. For these robotic fliers, it can be difficult to navigate in tight spaces without running into each other or environmental obstacles.

Now, a new machine learning algorithm is helping solve that challenge. Developed by researchers at Caltech, the Global-to-Local Safe Autonomy Synthesis (GLAS) artificial intelligence (AI) allows drone swarms to navigate crowded, unmapped environments.

Typical drone swarm navigation systems work by relying on existing maps of an area. For instance, a group of drones flying to light up the sky for New Years would have a dedicated map of the area above Times Square (or wherever it was flying). It also counts on knowing the route of every drone in the swarm. This helps each unit stay on track and avoid collisions.

The new GLAS program works differently. Thanks to its machine learning capabilities, it lets drones navigate an unknown area on their own while simultaneously communicating with other drones in the swarm. Using this decentralized model makes it possible for the drones to improvise.

It also makes it easier to scale the drone swarm. Since the computing power is spread across many robots, adding more of them is actually helpful.

While the main AI helps the drones navigate their environment, a secondary tracking controller called Neural-Swarm helps them compensate for aerodynamic interactions. This could be something like the downward force of air coming from a drone flying overhead. The system is more sophisticated than many controllers available today since they dont account for aerodynamics.

While drone swarms are almost exclusively used today for entertainment, that will change in the future. Drones are simply too helpful and too versatile to only be used for light shows. As such, the new swarm monitoring technology has plenty of applications.

While light shows will certainly benefit from it, other areas have even more to gain.

Search and rescue is a field that would certainly see improvements. Equipped with the new software, first responders could deploy a swarm of drones to quickly and effectively cover an area. This is only possible because the GLAS program allows them to operate without relying on a map of a predetermined area.

Meanwhile, the tech could be helpful in areas that dont include flying. For example, it could be integrated into self-driving cars to help avoid crashes and traffic jams. If every car included such technology, they would be able to autonomously interact with each other and adjust accordingly.

Despite these exciting applications, it will probably be several years before systems like GLAS are integrated into real-world drone swarms. Much more testing is needed to ensure it is reliable. Still, thanks to innovations like this, drone swarms will one day be commonplace.

Continued here:

AI is helping drone swarms fly in unknown locations - The Burn-In