Unmanned drones are risky business. As Wired notes, flying in the States, and particularly landing, is an extremely complicated process: Air Traffic Control is responsible for coordinating with airborne planes to negotiate landing patterns. If a landing strip is too busy, planes may have to circle while they wait, but that requires more coordination so the waiting planes don't smash into each other.
That kind of information processing is no trouble for a trained pilot, but unmanned drones can't reliably be trusted to react properly.
But drones also have their benefits: If that problem can be addressed, industries like shipping would be able to use a fleet of cheap, unmanned drones that follow a manned aircraft. FedEx, for example, is working on just such a fleet.
So it's no surprise that the Air Force Research Laboratory at the Wright-Patterson Air Force Base, just northeast of Dayton, Ohio, announced it will be looking for new ways to get unmanned drones up to speed. The submissions are to be algorithm-based; the Air Force hopes a well-designed algorithm, linked to the Air Traffic Control center, would be able to process more intricate information than current unmanned drones can handle.
These algorithms would allow a drone to recognize the "intent" of other aircraft, referring to its intended path and not just how it appears at the moment. Wired gives this example:
For instance: aircraft landing on parallel runways can appear to be on a collision course before they turn and land. Right now, a drone would simply perceive that a plane’s trajectory is going to remain unchanged, making it a threat for collision. But a capable algorithm would let the drone process Air Traffic Control information like basic airfield maps to know that there’s no actual danger from the oncoming piloted plane.
This post was originally published on Smartplanet.com