DARPA just wrapped up a year-long program to demonstrate that unmanned aircraft can perform in-flight refueling autonomously. The 10-flight series was flown by a NASA F/A-18, modified for “hands-off” flying, and a commercial Boeing 707 tanker.
Credit: NASA (link to giant-size photo)
This was the first time in history that autonomous refueling has been attempted, says DARPA. I spoke with Air Force Lt. Col. James McCormick at DARPA's Arlington, Va., headquarters on Aug. 17 about what the results of the test flights mean for the future of refueling.
Initially, the focus of the test was advanced unmanned combat aircraft, says McCormick. But applications for manned aircraft are just as important. There were 34 class A, B or C mishaps in fiscal 2006 and 2007, across the services. “Pilot skills are important to the mission,” says McCormick. “The ability of the pilot to put the little peg into the hole doesn’t contribute directly to the mission. If there were an ‘easy’ button that would reliably and repeatedly accomplish that, then the pilot could focus on more important things.”
Rather than using an Air Force-style boom-and-receptacle refueling system, DARPA chose a Navy-style probe-and-drogue system, which has the added challenge of unknown drogue motion relative to the tanker frame of reference (learn differences here). “The challenge was to use an optical tracker to take out that uncertainty and be able to track the less stable motion [of the drogue] when compared to a rigid boom,” McCormick says. The results were highly successful. Over the course of 8 tanker flights, a total of 97-plus attempts were conducted in a wide array of conditions and configurations. The number of plug attempts per flight went from about five in the first four flights to almost 20 in the last four.
There were a number of different algorithms that went into processing the GPS, the inertial positioning differences between the tanker and the aircraft, integrating an optical track and coupling into the flight controls. Corrections were made along the way. “In our effort to streamline the coding so it would run real-time,” says McCormick, “we oversimplified the translation of the tanker going into a turn. It caused a slight offset in the response. So we fixed it.”
For the most part, the test flights went well. Very well. “What surprised me was how effective the system was at the end,” says McCormick. “We switched between different software configurations in the last couple of tests. We had a good number of misses as well as effective plugs. One configuration consistently worked every time.”
The “exceptional performance” achieved during the tests was a result of two major enhancements to the AARD system, says DARPA. Improved video processing and advanced control algorithms, capable of anticipating and matching the drogue’s motion. “Some of the ways we anticipated [the motion of the drogue] actually made [the plugging-in process] worse,” says McCormick. “The most successful approach was specifically trying to anticipate the motion of the drogue itself, not due to [the F/A-18’s] bow wave.”
Will there be another set of tests performed with an unmanned aerial vehicle? “This program did what it set out to do, and that’s the end,” says McCormick. “Personally, I’d like to see DARPA or someone else pick it up and take it further.” Until that happens, the algorithms used in the AARD test would be useful in reducing workload and increasing safety for manned aircraft. “There is, of course, an integration challenge,” says McCormick. “Any time you touch the flight controls it’s open-heart surgery.”
Hopefully the services that have expressed interest in developing this program will do so openly, and soon. The results of the DARPA tests have not only demonstrated that autonomous refueling is possible, but that the technology isn’t out of reach and the cost isn’t prohibitive. How many other programs can you say that about?