Aviation Week & Space Technology

Podcast: Will Airplanes Ever Fly Themselves?

Discuss this Video 24

on Aug 19, 2016

"Self-flying aircraft are even further off than self-driving cars."

Really? For many purposes the problems are a lot simpler with more time to cope with identifying threats and to take necessary actions.

Moreover the navigation aids are better and the rules of the road are simpler.

Most systems and situations a pilot has to cope with are far more consistent than city streets and parking lots. There are few jay walkers (well, ok, an occasional Canada Goose), less idiots flying than idiot drivers. The kid with their head stuck in pokemon go isn't headed for the intersection when your light turns green.

I suspect that we will see self flying drones before we see self driving cars in wide use. The problem is actually simpler and the commercial path seems clearer.

Once upon a time folks thought developing the intercontinental cruise missile would be far easier than developing an ICBM.

Boy were they wrong. They could not see that the wings and things they were used to in airplanes made the intercontinental missile problem much harder.

on Aug 20, 2016

Good thinking... the complex unknowns in 2-dimensional road traffic far exceed 3-dimensional air traffic... it's a function of potential density within a relative future time frame.

on Aug 22, 2016

Agree, but its not only a matter of potential density. Aircraft go by reasonably identifyable plan, whereas road traffic is close to stochastic. And, even if aircraft plan (control reference data) were not exchanged, it may be guessed reasonably well. In fact almost all aircraft today are "self flying", the flight crew performing supervision and mode inputs for a large percentage of the time.

on Aug 22, 2016

Fully agree.
Time magazine actually ran a story on these lines a few years ago before autonomous cars became a fashion subject.
Also autonavigation (known-point to known-point) like on the cruise missile is far easier than autonomous (arbitrary point to arbitrary point) navigation. In autonomous the last few percent or percentiles are the most difficult in terms of robustness and testing. Autonavigation is sufficient for 98% of the use cases out there and far easier in development and maintenance.
And, in case of complete malfunction the fact that something which went up will come down makes the subsequent trajectory more predictable. In an autonavigation setting it can actually be precomputed along the complete trajectory. That makes development of mitigating safety measures easier than for ground based traffic.

on Aug 22, 2016

Excellent points. The only thing I'm happy with now about that statement is that it provoked a great discussion.

--Jim Asker
Executive Editor

on Sep 8, 2016

"Self-flying aircraft are even further off than self-driving cars." is a puzzling statement in a trade journal which has covered the evolution of autonomous drone aircraft since their inception.

By comparison, the recent collision of a car equipped with what's considered to be one of the more highly-advanced collision avoidance systems with a semi-trailer in Florida (resulting in the decapitation of the driver as the car drove under the trailer at speed) illustrates the opposite - that the problems of managing vehicle separation are closer to being solved in aircraft than they are in motor vehicles. Of course, no one flies in formations nearly as dense as rush-hour freeway traffic..

Not that I believe self-driving automobiles are an insoluble problem, but that self-flying aircraft are, by comparison, not so difficult to achieve. Fewer vehicles to separate, stricter laws regarding separation of aircraft, more room to separate them.

The scary scenario to me is the "aircar," which might have been technically feasible, but if the per-unit cost got down to, say, USD 100,000 or so we'd see air traffic density increase to the point where aircraft separation became as difficult as separating motor vehicles on freeways in dense traffic.

on Aug 22, 2016

Good solution for auto-eradication of human race when those things are hacked by terrorists and jumped onto the ground!

on Sep 8, 2016

There's no reason GPS couldn't be hacked in the same way to complicate precision approach landings at busy airports. Just a matter of using FM capture or other properties of radio signals to spoof or override navigational radio signals. It's probably an under-studied problem.

on Aug 22, 2016

What do you think drones are? Clearly it is technologically possible for all aircraft to be self flying. It is only a question of will. Will people want to fly in a computer controlled only aircraft. As more self driving cars are out I am pretty sure people will not have issues with computer controlled aircraft. After all fly-by wire means the pilot is flying the computer that is controlling the aircraft.

on Aug 22, 2016

A donkey is more reliable than an UCAV and can't be hacked!!!

on Aug 22, 2016

We will be removing pilots from the cockpit much sooner than most people think. AI is improving at an ever increasing rate annually, and when you factor in that the vast majority of airplane/airliner accidents through history have been due to pilot error, and the cost of employment for pilots to the airlines, I think it will be a no brainer for the switch in the 20-25 year timeframe.

on Sep 8, 2016

"No-brainer" sums the problem up pretty well. We do all right with aircraft autopilots for commercial air because we don't try to replace human pilots with them in airliners (where the problems of aircraft separation and landing are significant).

I said above that "Self-flying aircraft are even further off than self-driving cars" was a difficult statement to justify, and given the relative number of miles flown safely on autopilot in commercial air versus miles driven in self-driving cars, that was a accurate statement.

And if the relative traffic densities and current ability of sensors and guidance software to keep traffic separated remains the same between cars on freeways and aircraft in relatively uncongested airspace remains the same, it's still a true statement.

But let's say we get computers, sensor suites and software good enough to taxi an airliner from jetway to runway, rotate into take-off, and avoid collisions from take-off through the flight profile to landing, and taxi the aircraft back to the destination ramp, all unaided by human pilots.

As aalexandre pointed out, there be terrorists and just plain bored jerks with too little to do and too much hardware in their hands to try and cause accidents (consider the rash of 'lasing' incidents and then substitute microwave devices for lasers aimed at unmanned airliners while on approach to interfere with guidance signals, and you see the issue). Another issue is simple unintentional failure of guidance or engine control circuits (a problem with FADEC caused the crash during takeoff of that Airbus military transport in Spain).

While we may be closer to self-flying aircraft than self-driving cars, self-flying aircraft capable of replacing attentive and disciplined pilots (not guys like that Northwest Airlines flight crew who overshot their destination by 100 miles while doing stuff on their laptop computers, or the RAF pilot who brought a digital camera in the cockpit, where it jammed a sidestick flight control) aren't on the horizon yet.

I think it'll take more than 20 years for the entire flight profile for an airliner or heavy transport to be automated entirely, with no trained pilot in the cockpit and an acceptable degree of security against control failure resulting in loss of the aircraft and passengers. We ought to develop the technology, if only to allow remote recovery of control of hijacked airliners or aircraft whose flight crews have (for whatever reason) lost control of their aircraft.

Of course, that raises the issue of contention between pilots, automated flight control systems, and remote controllers - who gets the stick? The other issue will be politics. FAA's air traffic control infrastructure has, from time to time, been neglected. In the case of entirely automated aircraft, lapses in funding for that infrastructure are much worse issues than they were back when human pilots and air traffic controllers were responsible for safety issues. Take them out of the loop and we're making huge bets on how good the automation is.

on Aug 22, 2016

Almost as ludicrous idea as a square wheel! Truly 2/3 of the proficiency check I go through each year is to practice "failures" that occur in the airplane. Yes, there is redundancy in all important systems in the airplane but there needs to be a manager of resources to coordinate the operation. What about dealing with maintenance, dispatch decisions, flight attendants and crazy pax?! There may be a reduction to one pilot someday with far superior technology than we have today but there will "never" be a human-less commercial transport aircraft! What person would ever get on that airplane? Who would pay a high fare for that kind of unwarranted risk? Additionally, all car manufactures should cease and desist any idea of self driving cars. It is "human nature" that would make that idea sooo unsafe. If one does not have to pay attention to driving, they won't! They would be even more distracted and unsafe than they are now! How many times do you look over at another car and see them distracted with their smart phone or their tablet? JB

on Sep 8, 2016

You just made NHTSA's and the US Department of Transportation's case for self-driving cars - too many incompetent and impaired drivers out there.

Of course, you were comparing trained and licensed pilots whose health is monitored routinely and who have to undergo regular proficiency checks with automotive drivers. Politics have made the motor vehicle operator's license a substitute voter registration card, picture ID for every other purpose, everything but evidence of the ability to safely operate a motor vehicle. All you need in most states to be routinely granted renewal of a driver's license is the requisite fee.

The result is that 94% of all car crashes happen because a driver made a mistake. The nation's skies would look like the Battle of Britain if the current standards of driver training and proficiency were true of aircraft pilots.

No politician would be elected by promising to get all those unsafe drivers off the street, so self-driving cars are the reasonable compromise we'll reach, since no one in politics is gutty enough to tell the public "most of you need to retake drivers' ed, and you should ride buses until you're no longer a threat to other people's lives behind the wheel!"

That's the sort of talk that's kept Libertarians out of the White House for fifty years.

on Aug 22, 2016

Ha! The author of this piece is totally deluded. Self driving cars will be mainstream in 10 years, and self-flying planes even sooner. The weak link is the human link, and we won't be in driver's seat much longer. Get used to it,

on Sep 8, 2016

Tell that to the flight crew of the Airbus m400 Atlas military transport that crashed on takeoff from the factory strip in Spain. Bring a medium, because five of them are dead - the software for the Full Authority Digital Engine Control was misloaded, and they lost power on all but one engine.

If the flight crew had been aware of the issue (the software didn't foresee that happening in the ground and didn't alert the pilots) and able to fix it, those people might still be alive. Software can't replace pilots yet.

UCAV and similar autonomous drones have, for all practical purposes, the whole sky to themselves, and no passengers onboard, The technology's good, but it's not ready for the terminal airspace around a major airport yet.

on Aug 22, 2016

One of my friends called yesterday, he'd had a hard day at the office and wanted sympathy. The autopilot on his CRJ was u/s and he had to hand fly the thing through two trips before some other poor SOB had to take over. Normal procedure is to climb a couple of thousand feet then watch the autopilot fly it to a couple of hundred feet from landing. Even back in 1983 at RAE (Bedford) XX105 could fly itself from the start of the take-off roll to after landing, it couldn't use the brakes. It didn't because they weren't certain enough of the accuracy, but there were projects on other aircraft, like the HS748, that addressed blind landings without external assistance and with zero visibility. For the past many years my former boss has been working there on collaborative routing, aircraft discussing with others in their vicinity what their requirements and intentions are, then working out how to fit in. So this is all in hand. What isn't in hand is the super complex fault handling logic that is necessary when a compressor disk overspeeds and explodes through the wing and body and rolls the dice on every system. So while autopilot might soon make much of ATC and pilot roles redundant it can't eliminate the need for a couple of systems experts to fly the plane when the autopilot reaches the end of its logical tether.

on Aug 22, 2016

Airplanes today generally, already "fly" themselves. The correct question is whether airplanes will ever be allowed responsibility to plan and execute their own "mission(s). From receiving clearance to "pushback"; to acknowledging the instructions to take a particular taxiway off of a runway (after landing); and then hold on that taxiway, until crossing aircraft have passed; and, further instructions are provided.

In all three, planes, trains, and automobiles, there are two functional tasks and responsibilities. First, is the responsibility and task of the direction of the mission. The second, is the responsibility for the operation of the machine, and the execution of that operation (at the direction of the mission commander), in the fulfillment of the mission.

From intelligent cruise-control and automatic transmissions, to Airbus' flight director, these transportation machines of today, operate themselves, to configure; and reconfigure the machines, in order to fly or drive. They "assist" the vehicle commander, in continuous monitoring and configure/reconfiguring the mechanical state of the vehicle, to execute the commander's mission directions for transportation.

In fact today most air vehicles can't be operated directly, they are all, "fly-by-wire". So the question is really about that first role, the responsibility and task of commanding the mission. That responsibility is now a human one, solely. And the question is; is it now technically feasible that the responsibility for and execution of the command of a mission; can be entrusted to a non human; to plan; to execute; and to be responsible solely, for such mission's execution?

on Sep 8, 2016

Deconfliction of flight paths is a really thorny issue. Automating the various individual aircraft in an airspace doesn't solve it. Nor does it solve issues of unforeseen mechanical and electronic failure. There just aren't ballistic chutes big enough to save a heavy that just lost power with nobody in the cockpit to manage the issue. The Gimli Glider incident and Sullenberger's landing of that Airbus in the Hudson with no casualties ended well because awake and aware pilots were in control.

on Aug 23, 2016

The 2 experts in the potcast somehow missed, that the Airbus A-380 is capable of performing the vertical collision avoidance maneuver after a TCAS RA completly independent without any pilot input.


on Aug 23, 2016

You are correct - I neglected to mention the A380's autopilot-based TCAS ability. I also know that Honeywell is developing a system for its Primus Epic avionics that will allow the autopilot to nudge the aircraft in the proper direction before the pilot takes over (based on cases in which the pilot makes the wrong input). I guess my point was that these are the exceptions rather than the rule, which is interesting given automation's ability to react much more quickly (and probably more correctly) than the human.

on Aug 23, 2016

THE ANSWER IS: I already have a "self-flying airplane." I have had it more than two decades. I sit in the left front seat and it takes me where I want to go... I know...I am being a wise guy. But my point is, I no longer put up with being some peasant on any airline, in an airport being pushed around by some government dweeb that I would not hire to pull weeds in my yard. Yes, it costs more money to extricate oneself from the grasp of the bulbous bureaucracy; but isn't that worthwhile?

on Aug 24, 2016

The podcast talks about systems that are an extension of today's protection against stalls, or collision avoidance. If anyone thinks a passenger aircraft can fly without a pilot they need to speak to a mathematician/programmer. Fully debugging software, quickly becomes theoretically impossible by any known method even for simpler programs. You would have to be willing to accept a crash with the frequency of a desktop computer freezing up. I would rather go to Europe on a bicycle.

on Sep 27, 2016

The Air Force and NASA has Drones flying now in the USA as well as in the Middle East. The F4 Phantoms fly as QF4E Phantom Drones out of Tyndall AFB and Holloman AFB but the last one will fly as manneed QF4E out of Holloman on December 20 & 21, 2016. Then the F16A aircraft will assume the role of Drone aircraft currently flying out of Tyndall AFB and will fly out of Holloman AFB in 2017. The primary mission is to be used as target to be shot down by modern military fighter aircraft but a ground controller in Arizona/ Nevada or other locations have proven to be safe to fly unmanned Drones around the World with very little problem and getting better all the time. I don't believe that passenger aircraft will ever become totally pilotless and will need to maintain a qualified Aviator on board in case a problem surfaces to maintain safety. A majority of aircraft mishaps have been cause by Aviators not maintaining situational awareness or not fully understanding the on board systems and how they work so when a problem occurs they can easily trouble shoot and fix if fixable. Many passengers don't realize that commercial aircraft must be capable of flying on one engine before they enter service which makes them extremely safe.

Please or Register to post comments.

Sponsored Introduction Continue on to (or wait seconds) ×