Drones with Artificial Intelligence to improve autonomous flight in closed spaces


This new implementation could be key for autonomous cars, in addition to search and rescue.

Drone technology can be difficult to pilot in tight spaces and without being hit by obstacles. However, Caltech researchers are working on a way for drones to move quickly and safely indoors.

With a new automatic adaptation and learning algorithm, Global to Local Secure Autonomy Synthesis (GLAS), which allows swarms to navigate in crowded and unmapped environments. The system works by giving everyone a degree of independence that allows them to adapt to their changing environment.

Rather than relying on existing maps, GLAS makes each machine learn to navigate a given space by itself, even when coordinating with each other. This decentralized model helps drones improvise and makes it easy to scale the swarm as computing spans many robots.

This technology could be used in drone light shows, as well as more vital situations. But the most important thing is the search and rescue situations, where they can explore and minimize jams and collisions.