1 Lt Courtney Roepke, 21 AS C-17 Globemaster III pilot, performs pre-flight checks on a C-17 at Davis-Monthan AFB, Ariz. The aircraft and crew departed Travis AFB Oct. 10, 2018, and were staged at Davis-Monthan to support civil authorities during Hurricane Michael relief efforts. USAF Photo by MSgt Joseph Swafford
Roepke flies a C-17 after departing Travis AFB, Calif. USAF Photo by MSgt Joseph Swafford
USAF Photo by MSgt Joseph Swafford
Maj Nathan Masdon, 21 AS C-17 Globemaster III pilot, performs pre-flight checks on a C-17 at Travis AFB, Calif. USAF Photo by MSgt Joseph Swafford
By LT COL JASON HANSBERGER, 89th AIRLIFT WING Chief of Safety
Most of us assume we will have driverless cars long before we will be able to fly on a pilotless airplane. However, thanks in large part to a more favorable operating environment, pre-existing automation, and the nonlinear adaptation of new technology, pilots will likely go before drivers do.
Let’s start with the operating environment. The tech and auto industries are attempting to overcome the inherent mismatch between a computer’s ability to recall and crunch static data, and the dynamic environment where humans simply get in their cars and drive, responding to variables in real time. Conversely, the well-planned, low variable flight environment—with standardized arrivals, departures, and en-route clearances—plays well to a computer’s strengths.
Existing aircraft automation already meets the three primary capabilities required of a pilot: the physical capacity to fly; extensive knowledge of aircraft systems, procedures, and aviation rules; and sound judgment. When it comes to flying, no human can replicate the precision and accuracy of the autopilot. And unlike a computer’s capacity for data storage, pilots commit to memory only the routine information they encounter every day and the rarely encountered emergencies practiced in simulators and evaluated on check-rides. By contrast, as long as the data and programming is available, a computer is equally competent in the most routine and rare situations, requiring no training or currency for either.
Even without advanced artificial intelligence (AI), computers are already competing with humans on judgment. According to the Federal Aviation Administration and Boeing, approximately 80 percent of aviation accidents and close calls are a result of pilot error. Human factors such as fatigue, crew dynamics, poor decision-making, physiological problems (e.g., vertigo and visual illusions), and disregard for regulations, are common causal factors in aircraft accidents. Judgement errors are compounded by our limited number of sensors—eyes, ears, inner ears, and hands being the primary, yet fallible ones—and the time required by the brain to process the data presented through the aircraft’s flight instruments and warning displays into usable information. Computers, on the other hand, apply uniform and immediate judgement based on a multitude of independent and redundant flight and system sensors, follow checklists, and immediately cross-tell experiences and lessons systemwide.
Although far more people have died due to pilot error, events like the Miracle on the Hudson show that rare occasions will require some level of adaptability in an autonomous aircraft. However, the level of AI required to fly will be far less than that required to drive, and the airlines and manufacturers have terabytes of data based on millions of flight hours from which to teach the algorithms.
Predicting exactly when planes will fly without a pilot on board is difficult since people do not adapt to technology linearly. Adoption follows a geometric path with a few early adopters followed by a steep curve up. People used to think traveling faster than a horse would kill a human. However, railcars quickly dispelled that myth, and people adopted the technology overnight. So where are we on the autonomous curve? It is unlikely that flight engineers or navigators saw their demise coming more than 10 years before computers took their place, and pilots have probably already entered that window.
Yes, psychology and security are two conspicuous hurdles to passengers boarding autonomous aircraft, but concerns will ease as familiarity grows. Riding in driverless cars will knock down psychological barriers to flying without a pilot and vice versa. Cybersecurity also remains a concern, however, a computer is able to defend an aircraft from hackers more effectively than human pilots. For a pilot to override a hacker’s attempt at controlling the aircraft, he or she must both recognize the hack and, lacking a mechanical override, electronically disconnect computer control.
In the end, safety and economics will force the change. Although flight in autonomous aircraft will be safer than manned aircraft, the first move will likely come from the military and cargo companies that do not have to worry about the psychology of paying passengers. Economics will also motivate cargo companies, which can reduce the need for system redundancies and the heavy, expensive life support equipment keeping pilots alive. Passenger airlines will still require redundant systems and life support equipment, but autonomous flight will eliminate the need to pay pilot salaries, benefits, training costs, and travel expenses.
Flight attendants’ roles in emergencies should make them feel secure that it will be much longer before robots are moving down the aisles. But with an empty flight deck, they will be burdened with the duties of making destination weather announcements and appreciating you for choosing their airline.