Amazon delivers drone integration pitch

Amazon is among several companies and universities participating in the NASA Unmanned Aircraft System Traffic Management (UTM) research-and-development program, having flown its custom prototypes in the NASA UTM Technical Capability Level 2 campaign that concluded just over a month before AirVenture began. Multiple teams in locations around the country flew drones beyond visual line of sight over sparsely populated areas, all centrally monitored through the prototype UTM system.

Once the data is analyzed and results are fully evaluated, NASA and its partners in industry and academia will transition to flying drones beyond visual line of sight in more complex airspace, and over more densely populated areas. Those TCL-3 test flights will begin later this year or in 2018, so it will require some time before the drone Amazon brought to AirVenture is delivering shoes and toothpaste to impatient customers.

Amazon Prime Air officials dispatched to AirVenture included public relations specialist Harry Hartfield and Russell Williams, the director of manufacturing, integration and optimization, who also is a manned-aircraft pilot in his spare time.

“I fly a 1936 Porterfield,” Williams said, noting his airplane of choice is similar to a Piper Cub if a Cub had a round motor. It happens to be the type of aircraft UTM engineers might refer to as “non-cooperating,” since it never broadcasts its position via transponder or Automatic Dependent Surveillance-Broadcast (ADS-B), and thus poses arguably the greatest challenge to safe integration of airspace: Any drone flying autonomously, as Amazon intends its drones to do, will need to be able to detect and avoid it. “I care a lot about what the drone is going to do, sensing and avoiding other aircraft.”

Both Williams and Hartfield were a bit coy when it came to the specifics of the technology built into the drone on display, one of the larger and long-range drones in Amazon’s development fleet. When it comes to detecting other aircraft, the system will integrate with the UTM system and augment that with onboard sensors to detect aircraft like a 1936 Porterfield.

“Initially, that’s cameras, primarily,” Hartfield said. The onboard sensors feed data into flight control computers that utilize algorithms augmented by machine learning and “computer vision” to detect objects as small as a wire. Obstacle avoidance will be of prime importance (pardon the pun) throughout the flight, both during the transit (expected to be at or below 400 feet) and during the landing phase, when each aircraft must steer clear of any obstructions in the customer’s back yard to find the designated landing marker (deployed by the customer) on which to deliver the package itself.

“It can see things humans have a hard time seeing,” Hartfield noted of the system Amazon has developed.

He and Williams said visitors during the course of the week were “overwhelmingly positive.” Most of those visitors were pilots who will share the same airspace once the FAA allows package delivery drones to operate routinely, and Amazon is eager to gain support from the pilot community as one of the key stakeholders involved.

“I’ve been pleasantly surprised,” Hartfield said, though he did receive “a handful” of more negative reactions to the concept of autonomous drones navigating the low altitudes. Williams said the company will see to it the system does not pose a risk.

“We want to get our reliability to where we’re confident that the vehicle system is safe,” Williams said. There remains a great deal of work to be done on the regulatory side as well (an effort AOPA is directly involved in, through membership on the FAA Drone Advisory Committee). “When all these things line up, we’re ready to launch.”