Flocking ScanEagles

In military circles, it’s common these days to hear the virtues of unmanned aerial vehicles touted in Pentagon speeches and trade show pamphlets. According to the prophets, UAVs will soon be flying missions on their own, dodging aircraft, tracking terrorists, and sparing the U.S. Air Force the cost of recruiting and training actual pilots. But first someone has to develop the software to make the miracles come true. So when the people who program UAVs make a real breakthrough, it’s worth noting.

In late June, over the course of two days of flight demonstrations, a team from Boeing launched a trio of ScanEagle UAVs and, under the minimal guidance of a single human controller, let them fly missions on their own. The ground controller commanded the robotic craft to search for and track moving vehicles on the ground, then sat back to let them decide how best to accomplish their mission—including how to avoid being seen or heard by the targets.

ScanEagles, unlike joystick-controlled craft like Predators, already fly autonomously. Typically, one operator sets waypoints for the airplane to follow, while another controls the payload—a camera mounted on a turret. In June, a single operator controlled three aircraft simultaneously, taking the place of six people. He was able to do this because the aircraft were programmed to pick their own routes, divide up the terrain to be scanned, then find and follow targets without receiving any direct commands.

The demonstrations of the new software, called Distributed Information-Centralized Decision (DI-CD, or “diced”), were an unqualified success. That’s good news for Marshall Williams, the Boeing Advanced Systems program manager who taught this small flock of ScanEagles how to fly. Williams says it’s all about creating programs that respond to the way humans communicate, instead of training people to translate commands so computers can understand them. “It’s a whole new way of telling them what to do,” he says. “You tell them, I need pictures of those things in this space.”

In his more than 20 years at Boeing, Williams has worked on several software-intensive projects, including the F-35 Lightning, F-22 Raptor, and other secret programs he declines to identify. The June tests were the culmination of nearly four years of his team’s work on developing the software that runs UAVs.

“Many software engineers never get to see their technology flight tested,” he says. “I was able to take some of our young talent from the idea [stage] to the flight test…I was charged.”

On June 20, a group of company VIPs and customers (the identities of which Boeing declined to specify, citing industry competition) took their seats in a building at the company’s high-desert test facility in Boardman, Oregon. A mile away, in a mobile command unit, the three ScanEagles launched as they always do—catapulted into the air with specially designed trebuchets. Air Force operators waited nearby, ready to take control in case the software lost control of the aircraft.

Out in the desert, far below the ScanEagles, were several moving trucks. The ground operator, using an Advanced Warning and Control System (AWACS) control panel, gave each ScanEagle its basic instructions: Find and follow the first truck. While one ScanEagle followed the truck, a staffer playing the part of field observer used a cell phone to tell the AWACS operator of a new, time-sensitive target—a second truck.

Given the new priority, the software generated a new mission plan and flight paths for the three ScanEagles. The aircraft sailed over the second truck and fed video footage back to the AWACS controller, who sent it to the “field observer” on his cell phone. For kicks, the scenario ended with a call for a simulated F/A-18 air strike on the target.

The next day, a UAV operator showed that the new software could play nice with other applications. Everything was done according to current NATO practices and standards, with existing equipment. Compatibility with existing soft- and hardware is critical, says Williams. “Most of the pieces already exist,” he says. “They just need to be connected.”

Teamed with an optical tracking program called Stalker, DI-CD kept monitoring the truck as it made several abrupt turns, stops, and starts to evade detection. The software automatically adjusted the ScanEagles’ flight path to optimize viewing angles, while staying out of sight. A human operator would find it difficult or impossible to keep a ScanEagle fixed on a weaving vehicle while manually plugging in waypoints.

Advances like DI-CD are fulfilling the early promise of UAVs, says Daryl Davidson, executive director of the Association for Unmanned Vehicle Systems International. “When you put more than one of these in the air, that’s where you really multiply your capabilities,” he says. And with a single controller operating several aircraft at once, the cost of staffing and training comes down. Davidson says it’s a sign of things to come, inside and outside the Pentagon. “Fifteen years ago it was like pulling teeth to get the services to operate UAVs in theater.…Now customers are demanding as many as they can get.”

A key programming hurdle that still needs to be overcome is collision avoidance. Eventually, though, Davidson and other UAV believers say robotic sensors will be better at this job than the human eye. “The technology exists, just not in the right format, that will make unmanned aircraft safer than manned,” he says.

In the meantime, Williams and his Boeing team are trying to widen the variety of hardware that can work under one brain. He thinks a rotorcraft UAV, paired with a ScanEagle and some sort of watercraft, may be the next big demonstration. Either way, it will be a real-world test. “A lot of competitors show lots of PowerPoint charts and cool stuff in simulators,” he says. “It’s a lot harder to go and demonstrate it with flying assets.”