Would you travel in an aircraft that is fully automated? Or do you prefer the feeling that a human being is in control?
If you prefer the latter, you are not alone. But is this an irrational feeling? Let’s take a look at the world of autonomous aircraft of the present and future.
Why are planes not autonomous?
In short, they already are to a certain degree. But, before we answer this directly, it might be useful to take a look at the history of autonomous functions in aircraft.
In the very early days of aircraft, the pilot was required to be on high alert at all times.
His or her attention was constantly focused on flying, controlling, and monitoring the situation in and around the aircraft to ensure no impending danger was looming. Such activities often left pilots with little time to actually do anything else and kept them physically and mentally occupied for the entirety of the flight.
While this was not necessarily a huge problem for short flights of a few hours or so, as aircraft become larger, faster, and gained increasingly longer ranges, flight times became longer and longer. With all the best will in the world, it is very tiring for a human being to be subjected to high levels of awareness and concentration for long periods of time.
This is not only exhausting for the pilot but is also potentially very dangerous. A tired person can, and often will, make lapses of judgment that could end in disaster.
For this reason, autopilot functions were developed, and surprisingly early on in the history of flight. Believe it or not, but one of the first aircraft to ever be fitted with an autopilot system, albeit rudimentary by modern standards, was built by the Sperry Corporation in the early-1910s.
This system involved the connection of a gyroscopic heading indicator and attitude indicator to hydraulically operated elevators and rudder. The basic setup enabled the aircraft to fly straight and level on a preset compass bearing for extended periods of time without the pilot’s full attention. Such a simple device saved pilots a great deal of workload, as you can imagine.
Autopilot systems became ever more sophisticated over time, with the Royal Aircraft Establishment in the UK developing a more advanced system in the 1930s. Called the “pilot’s assistor”, this autopilot system used pneumatically-spun gyroscopes to actually operate the flight controls to a certain degree.
Further developments were made throughout the next few decades, with improved control algorithms, servomechanisms and even radio-navigation aids added, making it possible for aircraft to fly autonomously at night or in bad weather. Such systems became so sophisticated in fact, that as early as 1947, a U.S. Air Force C-53 was able to take off, cross the Atlantic, and land, all completely under autopilot control.
In more modern times, larger aircraft, and those with more than 20 passengers, are generally required, by law, to have some form of automation built-in. The level of automation does vary, but most provide what is called three-axis control for the pitch, roll, and yaw of the craft.
In most cases, modern autopilot systems are computerized and are able to make split-second changes or decisions from a large amount of data input from sensors and other monitoring equipment placed strategically around the aircraft. This will include data such as the plane’s current position, heading, altitude, speed, etc., and automatically adjust the “trim” of the aircraft accordingly when needed.
These autopilot systems will also be more than capable of climbing the aircraft after takeoff, maintaining cruise control and level flight, and handling descent, approach, and increasingly, the final landing phases of a flight. Taxiing before takeoff, the actual landing, and taxiing after landing are still usually the preserve of the human pilots.
Autopilot systems will also generally disengage during periods of extreme turbulence, when the pilot(s) are expected to manually take control.
The same is usually also true for taking off. At least for now.
Airbus announced back in 2020 that it has successfully designed and tested a completely autonomous takeoff system. Forming part of its highly ambitious Autonomous Taxi, Take-Off, and Landing (ATTOL) project, this news was groundbreaking for the industry.
The technology employed differs from existing Instrument Landing Systems (ILS) that are common on modern airliners, and used image recognition technology to keep the craft on the runways center line, adjust trim and speed, and finally lift the Airbus test aircraft into the air.
This is an important step towards making aircraft fully autonomous in the not too distant future.
When will pilots be made redundant, if ever?
So, with the high level of sophistication of modern autopilots, you might be forgiven fo
r thinking that pilots seem to be something of an indulgence in the cockpit. If a plane can, in theory, fly itself, what need do we actually have for pilots?
As it turns out, while much of the actual heavy lifting, so to speak, of flying, can be delegated to a plane’s autopilot, a human presence is still extremely valuable. In fact, this is highly unlikely to change anytime soon.
One of the main reasons, for better or worse, is the general mood of the public, and of the passengers of the aircraft. Whether you admit it or not, there is something very reassuring knowing that a living, breathing person is in control of something like an aircraft (or at least at present). Most people are reluctant to give full control of something that could, theoretically, kill them, entirely over to a machine.
Would you, in your heart of hearts, actually trust your life to a computer? For most the answer is no, and understandably. After all, what if something went wrong with the autopilot and there was no one to fly the plane? What if it were hacked? What if the plane needs to make an emergency landing in an unplanned location, such as on the Hudson River?
Interestingly though, some recent investigations might show that the general mood among the public is changing on this, at least where some vehicles are concerned. Some opinion polls taken back in 2019 revealed that seven out of ten consumers believe autonomous cars drive better than those operated by human drivers.
The survey was conducted by ANSYS and polled more than 22,000 people in Benelux, China, France, Germany, India, Italy, Japan, Spain, Sweden, the UK, and the U.S. Granted this poll was primarily about driverless cars, but it seems people are getting more comfortable with the idea of trusting non-human operated transportation.
When it comes to some future theoretical pilotless aircraft, most respondents were not entirely resistant to the idea, but would rather wait until the technology is more mature.
“Although the public shares some safety concerns about both modes of transportation, many of those surveyed trust the technology that operates these vehicles. 71% of respondents believe that autonomous cars are better drivers than humans or will surpass human abilities within 10 years. 70% of consumers were comfortable with flying in an autonomous plane in their lifetime, the majority of which were not aware of the level of autonomy already involved in commercial flight,” wrote ANSYS in the report.
Another major reason that pilots will likely remain a feature of most flights, is that under certain circumstances, human beings are actually better decision-making machines than actual machines. For all the complexity of a computer, it still pales in comparison to the squishy and complex computer between your ears.
Your brain can take in enormous amounts of information at any one time, make quick decisions, and is able to improvise on the fly, so to speak. This kind of flexibility is incredibly difficult to replicate in a machine, if indeed it can ever be possible.
A human being also has access to some other sensors not available to a machine — senses. While less reliable than digital or mechanical sensors under certain circumstances, there are times when something that doesn’t sound or feel right to the pilot could indicate some issue that the onboard flight management system may not be able to detect. However, it is important to note that most pilots are trained to trust flight controls, readouts, etc. over their senses, as humans can also often reach false conclusions.
Modern flight management systems often, for example, rely on air pressure to calculate speed, angle of attack, etc. Should this system fail, for whatever reason, a professional pilot is trained and drilled to detect a potential fault and make manual corrections.
In some countries, or in some instances, GPS systems may be jammed for security reasons over sensitive parts of the country. Most modern aircraft rely heavily on GPS information for flight accuracy, and cutting an aircraft’s autopilot systems off in this manner makes a human pilot all the more important. Human beings do not need to rely on digital signals for navigation and are able to improvise or find other reliable means of navigation lost to the aircraft.
Not only that, but given the highly chaotic environment of flight, the unknown or very rare event can, and will, crop up from time to time. These are not situations in which a regulated and regimented machine is able to make decisions. This is especially true in a situation that has never been experienced before.
If you are still skeptical, a historic example might help convince you.
In 2010, a Qan
tas airliner jet with 450 passengers suffered a severe malfunction mid-flight. An uncontained engine rotor failure sent shrapnel throughout the plane, damaging several critical aircraft systems, including the aircraft’s landing gear. The onboard flight management system was overloaded with emergency errors and messages, which were impossible to deal with all at the same time. The pilots on station (as well as some off-duty pilots among the passengers) were able to improvise and successfully land the aircraft.
While it is conceivable an autopilot system may have found a way to do the same, it was the quick thinking and ability to improvise by the human crew that saved hundreds of lives that day.
Can planes be made fully autonomous?
As we have already stated, to a certain extent many modern aircraft are already partially autonomous. While normally only available, in a true sense, on large commercial aircraft, autonomous systems are also incredibly important in many modern military aircraft.
The F-16, for example, is one prime example. Aerodynamically speaking, is designed to be slightly unstable through a principle known as relaxed static stability (RSS). When disturbed in pitch or roll, an aircraft with RSS cannot continue to pitch or roll in the direction of the disturbance. Unlike the inverse, positive static stability, planes with RSS do not return to a straight and level flight if the pilot releases the controls.
This gives better cruise performance, reduced stick force to maintain a new speed without retrimming, and makes the plane more responsive mid-flight if the pilot needs to make dramatic changes to the aircraft’s position in 3D space, as less time is lost attempting to “override” an attempted correction by the airframe. However, this is obviously potentially pretty dangerous if the pilot is out of action for whatever reason.
For this reason, F-16’s, and some other planes with RSS, make use of something called a quadruplex, or four-channel, fly-by-wire flight control system to augment a pilot’s input from stick and rudder controls to produce the desired outcome in the position of the aircraft without inducing potential catastrophic loss of control due to the aircraft’s RSS.
The reason for this is to provide the aircraft with very high levels of maneuverability in flight. All of this occurs without the pilot’s input, and should this system fail, the aircraft will become nothing more than an expensive, speedy chunk of metal at the whims of gravity.
In effect, “you don’t fly an F-16, it flies you!”.
Other modern aircraft also take advantage of RSS, like the famous F-117 “Nighthawk” stealth fighter. This aircraft’s unconventional shape was designed to reduce its radar signature but came at the expense of the aircraft’s stability in flight. Like the F-16, the F-117 uses its own sophisticated fly-by-wire system to keep it on course.
For smaller civilian aircraft, some notable advances have been made in recent years towards making them autonomous, too. For example, back in 2019, the Technical University of Munich (TUM), along with its research partners, made something of a breakthrough in providing a means of allowing light commercial aircraft the ability to land without a pilot’s input or ground-based support.
The project, called “C2Land”, used a specially designed optical system and fly-by-wire system to land a small aircraft completely autonomously on a runway. This is a huge breakthrough, and with more testing and refinement, could prove revolutionary for the industry.
But, these are not really what people mean by autonomous aircraft — i.e. no human pilot at all. There are plenty of examples of large and small aircraft that are, ostensibly, fully autonomous currently in operation and under development — chief among them drones.
Most commercial drones today will have a high level of automation, like impact or accident collision prevention of some kind. They will also have some other automated features to constantly adjust power input and trim to allow them to hover. But even these drones will typically require human input over the remote control to “decide” where it goes.
There are some security drones that can work autonomously once an intruder has been detected. These drones launch on their own, get into position and take photos or videos of the intruder — all without any human operator (though controls can be overridden).
In December of 2020, drone company American Robotics won FAA approval for some of their fully automated drones. Developed by a Massachusetts-based company, these commercial drones will be used to deliver stuff without human input of any kind during flight.
The significance of this event cannot be understated, but it does come with some pretty important caveats. For example, the ruling limits the drones to operate in rural areas only, and below altitudes of 400 feet (122 meters), and they will have to have a maximum takeoff weight of 20 pounds (9 kilograms).
This is very interesting, but these kinds of drones could never be used to transport people around, like aircraft. So what, if any, potentially fully autonomous aircraft will soon be available?
One notable example is currently under development by a company called Joby Aviation. An all-electric vertical takeoff and landing (eVTOL) vehicle, this company may become the first to receive Federal Aviation Administration (FAA) certification for an autonomous craft.
The diminutive craft is being designed as a small air taxi and has recently agreed to something called a G-1 certification basis with the FAA. This is ba
sed on the company meeting the certification requirements for small planes under 19 seats and 19,000 pounds (8,618kg) of maximum takeoff weight.
Like the approval given for autonomous drones that preceded it, this is another important step towards making fully autonomous aircraft an actual reality.
Another interesting development in the area of autonomous flying craft are proposals for purpose-built airports for such craft, rather than using existing infrastructure. One notable example is a planned e-plane airport in the UK.
The proposal has been made by a British-based startup called Urban Airport, which has partnered with Hyundai to design and perhaps actually build a first-of-its-kind eVTOL airport.
Called the “Urban Airport”, the new project aims to be zero-emissions and provide all the infrastructure eVTOLs, and their passengers will ever need. This futuristic airport will be 60% smaller than a traditional heliport, and could, in theory, be installed in a matter of days or weeks.
It could also be moved to alternative sites if needed.
In the meantime, Hyundai stated it has plans to create its own eVTOL aircraft and is supporting the development of Urban Airport’s Air One hub as part of its plan to commercialize its aircraft by 2028. The new micro-airport is being built in the city of Coventry and will be ready by November 2021. The project won a 1.2-million-pound ($1.65-million) UK government grant. It will act as a hub for future eVTOL aircraft, such as air taxis and autonomous delivery drones.
Will autonomous planes be safe?
We’ve already touched on some of the safety aspects of autonomous systems above. But, a better question is: Will autonomous planes be safer than piloted ones?
To answer this, it might be worth spending a little time exploring the safety of aircraft today. For example, is there any truth behind the statement that flying is “one of the safest ways to travel?”.
As it turns out, it really is. Since around 1970, the number of accidents involving commercial aircraft (aircraft with more than 19 passengers onboard) has dropped gradually. This occurred while the number of flights has increased rapidly. By 2019, fatal accidents per million flights had decreased 12 fold compared to 1970, from 6.35 to 0.51 fatal accidents per million flights.
The reason for this is a combination of technological improvements to aircraft, stricter regulation on the aviation industry, and, of course, increased improvement in autopilot reliability and capability. In 1972, there were not one, not two fatal crashes that occurred when the crew were distracted by chatting (about football and Christmas plans)
From 2019 statistics, which is the last “normal” year of flight volumes, pre-pandemic, your chances of actually dying in a plane crash were practically zero.
In the United States, for example, the total number of accidents (not fatalities) involving U.S.-registered civilian aircraft (both big and small planes) was around 1,302 in 2019. This is down from 1,347 in 2018 and follows a general decline over time. There are some spikes in the data, of course, most of which relate to unprecedented events like the 9/11 terrorist attacks in the U.S. that obviously skew the data.
The number of civil aviation deaths increased from 395 in 2018 to 452 in 2019. All but 8 of the 452 deaths in 2019 were onboard fatalities (events like passenger heart attacks, etc). Four of these deaths involved a commercial airline.
While hundreds of deaths involving aircraft might seem like a very high figure, it is important to put this in perspective. Taken as a ratio per 100,000,000 passenger miles, your chances of dying due to flying is basically a rounding error to zero.
Compare this with passenger ground vehicle statistics for the same period. Accordingly to the National Safety Council, “over the last 10 years, passenger vehicle death rate per 100,000,000 passenger miles was over 9 times higher than for buses, 17 times higher than for passenger trains, and 1,606 times higher than for scheduled airlines.”
That is a huge difference, and remember these aircraft, and most vehicles, are still human-operated. But what difference, if any, would make the aircraft fully autonomous make?
In a roundabout sort of way, the safety of larger commercial aircraft is sort of a proxy for this potential future of flight. The FAA, for example, estimates that most aircraft are flown on autopilot about 90% of the time.
But, as we’ve detailed before, autopilots can, and do, make mistakes, too. It is for this reason that human pilots are still required in cockpits to act as a sort of “sense check” or emergency backup should the unthinkable happen.
Not only that, any automated or computerized system is only as good as the programming behind it, and the proper use of it.
Logical in theory, but organizations like the FAA have raised some concerns that the reliance on autopilots is making pilots lose their edge.
According to a report prepared by the FAA, “several recent accidents, including the July 2013 crash of Asiana Airlines fligh
t 214, have shown that pilots who typically fly with automation can make errors when confronted with an unexpected event or transitioning to manual flying.” Industry experts’ concern is growing about “whether pilots are provided enough training and experience to maintain manual flying proficiency.”
Most airlines do encourage their pilots to fly manually when safe to do, and as much as possible, but the entire point of such autopilot systems is to reduce stress and fatigue on pilots and are used accordingly.
But how often do autopilot systems actually fail?
Sadly, there are no reliable statistics on this question, but several decades of widespread use of autopilot systems seems to indicate that failures are very rare, or are often the result of user error.
But, we do have some data from autopilot systems in other modes of transport, like cars. Much like similar systems in aircraft, autonomous functions in cars have shown a dramatic reduction in accident rates in cars.
Some studies, for example, are showing that autonomous vehicles are already drastically outperforming human-driven vehicles in terms of safety. According to the Department of Transportation (DOT) and the National Highway Traffic Safety Administration (NHTSA), almost 94% of accidents on US roads occur due to human error, so self-driving vehicles could drastically reduce the number of crashes and fatalities that occur on the roads today.
And there are still many accidents every year on the roads. In 2019 alone, just over 36,0000 people were killed in road traffic accidents in the United States. The reasons for these crashes differ widely, but driving under the influence of drugs or alcohol, not paying full attention for whatever reason, or driver fatigue.
Since an onboard automated driver will never drink, cannot be distracted, and won’t tire as long as it has power, widespread rollout of this technology should result in a dramatic reduction in this figure over time. They also make significant improvements to fuel efficiency too. Autonomous systems in vehicles are typically able to find the shortest and/or fastest route between two points.
But, autonomous cars are still pretty rare so making a realistic assessment of the impact on road safety is very tricky. However, some organizations, like Google, have conducted some pretty in-depth studies on just this subject.
Through its Google self-driving car project, they found that autonomous vehicles were involved in 13 fender benders, and no serious crashes, over a total of 1.8 million miles of autonomous driving. That is impressive, and perhaps an insight into the kind of impact this technology can make on our roads.
However, it is important to note that autonomous technology in vehicles, while impressive, is still in its infancy and is not immune from making mistakes. Just like pilots, human drivers, so long as they are compus mentis, tend to be more adaptable to rogue situations than pre-programmed computers.
This is changing though, as machine learning is becoming an ever more important component of such systems. Given enough driving experience, and time, these systems should be able to match, if not excel, in this area too.
So, to summarize, how close are we to fully autonomous flight?
Technologically, we seem to be more-or-less there. Once autopilot systems can be developed, tested, and trusted by organizations like the FAA, the role of human pilots will continue to dwindle over time.
However, just like today, we will probably never see a future where no trained human being is present in the cockpit for commercial aircraft. Even if we assume that all the technical kinks are worked out, and autopilots become adaptable as human beings, many passengers will likely feel safer knowing a human is present and ostensibly in control.
However, when it comes to delivery drones, military drones, and perhaps even military aircraft, a pilotless, fully autonomous future is likely an inevitability.
Whether or not that will change over time, especially as the general public becomes more accustomed to autonomous vehicles, is yet to be seen. But, what is clear, is that further automation in all aspects of transportation (planes, trains, and automobiles) is likely going to make travel in the future even safer.
So long, that is, that these systems are safe, unhackable, and protected from massive electromagnetic attacks.