Input/Output: Learning from UAV Mishaps

The explosion of interest in unmanned air vehicles (UAVs) has focused attention on their high accident rate--several orders of magnitude greater than that of manned aircraft. The U.S. Defense Department's 2003 "UAV Reliability Study" reported that human error was involved in 17 percent of UAV mishaps, but other studies have put the figure between 34 percent and 51 percent.

A U.S. Air Force study completed in March of 2005--"U.S. Military Unmanned Aerial Vehicle Mishaps: Assessment of the Role of Human Factors Using HFACS"--examined 221 UAV accidents in the U.S. Air Force, Army, Navy and Marine Corps over a period from 1994 through 2003. (The study found that human factors were present in 133 accidents, or 60.2 percent of the total sample.) The study reassessed existing data, using the now standard human factors analysis and classification system (HFACS). In the USAF study, "mishap" implies damages ranging from more than $20,000 through complete destruction of the air vehicle.

The study was undertaken by the USAF's now 8-month-old Performance Enhancement Directorate. Its purpose was not necessarily to address specific issues, but to establish an empirical assessment of UAV human factors issues as a baseline for going forward, says Maj. Anthony Tvaryanas, chief of the directorate's UAV Human Systems Integration Branch and one of the report's authors. The study did raise isses that need to be studied further.

The differences between the services' approaches to UAVs and mishap investigation led the authors to partition their data by service. At this time, for example, the Air Force employs trained pilots to operate large vehicles such as Predator and Global Hawk. The Army, Navy and Marine Corps, by contrast, typically operate smaller, less expensive tactical UAVs, generally controlled by enlisted personnel who are not pilots.

Nevertheless certain factors are common. "We need a thorough human factors evaluation of the GCS [ground control station] environment to help develop UAV-focused HCI [human-computer interface] design rules and principles," Tvaryanas says. "We need to evolve standards for component reliability, training and interfaces, just as there are for manned aircraft." Some Air Force mishaps have been associated with instrumentation, sensory feedback systems and automation, resulting in channelized attention, where the operator is so focused on one thing that he neglects others. Fifteen Predator mishaps have involved HCI issues. And, although operators in Class A airspace are legally required to be instrument-rated pilots, there is room for debate about how best and most cost-effectively to train them and how they should do their work. UAV operators lack the rich sensory environment of manned aircraft, including peripheral vision, auditory and haptic cues. The USAF report cites a NASA summary of lessons learned, indicating that providing UAV ambient noise to the GCS "proved invaluable and potentially saved the UAVs in some cases." Should future systems use nonvisual cues?

NASA's report also recommended limiting or eliminating multifunction switches on the HCI and making "the status of critical parameters...easily observable." Predator's current GCS is "heavily reliant on multifunction keys driving a hierarchical system of computer windows," the USAF report says.

The USAF study also reported surprises. Air Force operators were found to have a higher proportion of mishaps attributable to skill-based errors. These errors involve "highly automotized psychomotor behaviors that occur without significant thought," such as classic stick-and-rudder skills. Most of them were procedural errors.

The report cites an Air Force Research Lab study indicating that 150 to 200 hours of flight time was enough for most pilots to achieve the stick-and-rudder skills to perform basic Predator maneuvers and landings. It also says that experienced pilots did not significantly outperform less experienced pilots on simulated, basic UAV flight tasks. And it cites evidence suggesting that experienced pilots may need to unlearn certain aspects of piloting, such as dependence on vestibular and peripheral vision cueing, especially during landings.

In addition, the USAF study reported that experienced military pilots/operators made as many judgment errors as enlisted operators without prior military flight training or experience. And it found no difference between services in the frequency of mishaps involving crew resource management.