8 Technology Quarterly |Aviation The EconomistJune 1st 2019
1L
ast yearBoeing’s Pilot Outlook report estimated that civil avia-
tion will require 790,000 new commercial pilots over the next
two decades. Of those, 261,000 will be needed in Asia and 206,000
in North America. One approach to this problem is to open more
flight schools. An alternative is to need fewer pilots. And that re-
quires better technology.
The first autopilot was invented surprisingly early in the his-
tory of aviation, in 1912, less than a decade after the Wright broth-
ers’ original flight. It used a gyroscope and altimeter to operate a
plane’s control surfaces to keep it flying straight and level. Since
then, autopilots have evolved into flight-management systems
that can run almost every part of an aeroplane’s journey except
taxiing and take-off, and even those are starting to come under
automatic control. As recent events have shown, flight-manage-
ment systems are still not good enough to be trusted completely
when lives are at stake. But in a world where automated drones
such as America’s Global Hawk reconnaissance vehicle routinely
fly military missions, and self-driving vehicles are talked of as if
they were just around the corner, the question of how large a civil-
ian flight crew needs to be is clearly open for debate.
It is also a pertinent question for the armed forces. In particu-
lar, pilotless aircraft can be sent on missions too dangerous for
people, and possibly ones that piloted craft would be incapable of
performing. This could change how future wars are fought.
Both Airbus and Boeing are preparing for at least a single-pilot
commercial-aviation world. Such a world will require not only re-
liable flight-management systems, but also a redesign of the cock-
pit for one-person operation. Both firms are now testing simula-
tors of such cockpits. Airlines are keen. A report published last
year by ubs, a bank, suggested that moving to single-pilot opera-
tion could save the world’s civil-aviation companies $15bn a year.
Going fully pilotless would increase that figure to $35bn.
Moving to single-pilot operation would require an aircraft’s
flight-management system to be good enough to take over in a
medical emergency that incapacitated a lone aircrew, flying the
plane to a nearby airport and landing it safely. For all practical pur-
poses that is true already. It is only during take-off that the human
touch is still widely regarded as necessary. Pilots’ unions say they
are worried about the ability of a single pilot to handle an emer-
gency brought about by a problem with the aircraft itself, such as
an engine failure. But how necessary it is to have two crew mem-
bers to deal with such exceptional circumstances is moot.Cargo flies itself
Even so, automated flight systems are an area where the innate
conservatism of aviation technology manifests itself. For example,
flight-control software for civil aviation cannot easily take advan-
tage of the “deep learning” capabilities of artificial intelligence.
Regulators are loth to license anything they cannot understand.
But the whole point of deep learning is that it reprograms itself in
unpredictable ways in response to circumstances. That would be
of huge benefit, for it would permit the autopilots of individual
planes to learn from each other’s experiences.
In civil aviation, then, passenger aircraft are likely to remain
twin-crewed machines for some time. But that may not be true of
freighters. There would be no customer resistance here either to
single-pilot operation or even to full dronification, as soon as thatis possible. And if pilotless freighters proved safe, in a world in
which self-driving cars had also become commonplace, passen-
gers’ attitudes might change.
In America’s armed forces the next set of aircraft likely to drop
the pilot is helicopters. Much of the work is being conducted under
the aegis of the Defence Advanced Research Projects Agency, or
darpa, a research arm of America’s defence department. As part of
its alias(Aircrew Labour In-Cockpit Automation System) project
the Sikorsky Aircraft Corporation, which is the helicopter division
of Lockheed Martin, has refitted one of the firm’s s76-bcommer-
cial models to be able almost to pilot itself.
As Chris Van Buiten, vice-president of innovations at Sikorsky,
observes, flying a helicopter is far harder than flying a plane. Heli-
copters are aerodynamically unstable, so simply keeping them
straight and level is a challenge. But the whole point of a helicopter
is that it does not just fly straight and level. It can dodge around to
avoid obstacles, meaning it can hug the ground—all of which adds
to the mental effort of controlling it. And it is often deployed in
weather that would make a fixed-wing pilot think twice.
matrix, as the company dubs its experimental helicopter co-
pilot, has a central processor that receives signals from a range of
sensors, and combines these with data from the Global Position-
ing System and a map of the local terrain stored in its memory
banks. The processor then sends appropriate signals to actuators
located in various places around the craft’s airframe, to control its
mechanical systems. The mission-instructions themselves come
from matrix’s human commander, via a tablet computer, but Mr
Van Buiten hopes that voice-recognition systems will soon be-
come reliable enough for those instructions to be spoken.
One of the advantages of matrixis that it can respond much
faster than a human pilot. This is valuable in normal circum-
stances. In an emergency, it may be crucial. It can, for example, re-
act to an engine failure, assess a score of options for a forced land-
ing, and recognise which is safest, all within a hundredth of a
second. A human pilot would need a couple of seconds just to work
out what was going on.
The next step on its journey, which should happen in the au-Drop the pilot
There will be resistance, but crewless planes are on their wayAvionics