A research group at Caltech has been working on an artificial intelligence system that will give drones the ability to fly together in tight spaces without hitting each other (or other objects). Called Global-to-Local Safe Autonomy Synthesis (GLAS), the system uses machine learning to give each drone the ability to move independently of its counterparts while maintaining awareness of their location. For example, if one drone has to move to the right to avoid a pillar, the drones on either side of it don’t have to move with it and could even move away if it comes toward them.
Because this system uses AI and machine learning, it doesn’t require a detailed map of an area to be preprogrammed into the drones. This could be useful in situations like search and rescue, allowing multiple drones to be released immediately to search an area without having to take the time to give them a map first. The team also envisions this technology being useful for other automated systems, like self-driving cars in a traffic jam.
