Aviation Week & Space Technology

Podcast: Can A Robot Be Your Co-pilot?

Discuss this Video 26

on Nov 18, 2016

Better get used to the idea of a robot co-pilot.

In fact soon after self-driving cars become common the hard sell will be that self-flying airliners will be safer than ones with fallible pilots.

The unemployment line starts back there after all the truck and taxi drivers.

on Nov 30, 2016

Doubt it

on Nov 18, 2016

With the global shortage of qualified airline pilots approaching, there is no better time to get serious about reassigning to new technology certain roles traditionally associated with First Officer. Doing so would allow captains to fly with a reduced workload, no longer needing to direct finite energies toward deriving an acceptable level of "CRM" out of flying in an environment of absurdly asymmetric experience levels (extremely common outside the USA nowadays). CRM could mostly be replaced by ARM, "Airplane Resource Management," and First Officers' roles reduced to generally routine tasks such as handling communications. The crew composition and duties on the new flight deck would resemble that of two-man fighter and attack aircraft of yesteryear, where both crew members were trained to do an important job but only one was actually a pilot. A captain-qualified relief pilot would still be carried for longer duty periods; robotics would communicate via satellite and fully take control in the event of pilot incapacitation or deliberate sabotage; but with a moderate increase in the implementation of new technologies, the era of two heads trying to operate as one could finally come to an end.

on Nov 18, 2016

I'll have to ask Sully how much a robotic copilot cound have "helped " him when he decided that landing in the Hudson River was the best option.

on Nov 18, 2016

Don't think of it though in regards to today's autopilots and technology. If you recall, during AWE1549, the co-pilot spent a decent amount of the 180 seconds they had running through engine restart procedures. With the AI of the near future, an automated system would probably have been able to determine in less than a second that engine restart was impossible which in turn might have provided enough decision time for a safe return glide to LGA.

on Nov 18, 2016

As a pilot of a single pilot jet, the autopilot already is your co-pilot and the pilot, me, is just a systems manager. It works perfectly so long as it works perfectly. We are losing our hand flying skills and it is showing in the accident data. The next generation will be severely disadvantaged because the aircraft they are flying are also autopilot driven and they will not have the years of hand flying experience that old guys like me have. Good luck!

on Nov 18, 2016

That's one side of the story. The other side is that the vast majority of aircraft accidents have been caused by pilot error. Both in action, and also in failure to recognize and correct.

on Nov 21, 2016

The autopilot system would only be as good as the data programmed into it, thus any situation outside it's programming would likely result a situation being mishandled. In the future it will be interesting when 100% autopiloted aircraft start crashing (and they will) to see how many accidents are attributed to 'system error'. Though of course you will have at that time eliminated pilot error...

on Nov 18, 2016

There are lots of questions raised in this fascinating podcast:
* Does the automatic system need to be as good as pilots for failures that are not programmed into the system, such as uncontained engine failures or malicious damage?
* In my professional life, it was common for the engineers to not adequately understand flying and for the pilots to not adequately understand engineering. Sometimes this was resolved satisfactorily, sometimes not. To the degree that is true, who's going to develop such systems?
* Similarly, what happens when managers and bean counters step in and mandate an immature system due to schedule or budget pressures? There may be no human pilot resources available to take up the slack for a poor design, as is the case now.
* Many of the "pilot errors" occurring these days arguably have excessive complexity as a root cause, in airspace design, airframe limitations, operational constraints, warning messages, etc. Would it not make sense to simplify the operating realm rather than to try to master all the special cases?
* Ideally, this system would simplify training and reduce costs, but if this is an add-on system to existing complexity, that seems unlikely. And if the single pilot is having a bad day or forgets something or succumbs to some cognitive error...
* Along those lines, cockpit user interfaces are frequently flawed, with user interface fundamentals sometimes ignored. Similarly, mode confusion and "what's it doing now" questions abound. Is the solution then to add more automation, more gasoline to the fire? And tablet user interfaces can quickly become too deep to learn or to use.
* How is product liability to be addressed? The old rule of thumb was that the cost of developing a new system should include the cost of two product liability trials.
* How do you certify all this? Is rule-based programming ready to make a comeback?
* And there are of course more issues than just these.

The Airbus philosophy of preventing the pilots from doing dumb things is an interesting antecedent to study, for both good and bad.

Fascinating questions for single pilot flight. If I weren't retired, I'd sure like to put in a decade or so on this problem.

on Nov 18, 2016


on Nov 18, 2016

I foresee a future in which it will be illegal for direct human control of flying vehicles (and, for that matter automobiles).

I am not the first to say this; one might read some of the works of science fiction writer Larry Niven.

Friend Ed, above puts things very clearly in perspective- Well said!

on Nov 18, 2016

Read Chesley Sullenberger's book, "Highest Duty" to learn why years of experience and a passionate focus on safety are what helped save the lives of all onboard his aircraft. Lack of hands on flying due to automation isn't so good. Seems like a Korean airliner had that experience coming into SFO just a tad bit low & slow while depending on automated equipment.

on Nov 18, 2016

Sorry to nit pick, but the problem was that they had switched the automation off. They had started out too high and fast at 5nm, then made it worse by selecting FLCH mode with a higher altitude set, which caused the aircraft to climb. The pilot then disconnected the autopilot and brought the throttles manually back to idle, which also killed the autothrottle. Having taken over full manual control the pilot flying then expected the aircraft to realize he was expecting it to take over throttle control again, and the rest of the crew were expecting the very senior pilot to rescue the situation and kept quite until it was too late. They passed through the glideslope at the correct speed, but kept descending fast and slowing. By the time people started complaining it was too late to avoid the crash, they initiated the go-around at 100 feet with the engines at flight idle and no hope of getting enough power and speed to halt the sink before striking the seawall.

on Nov 18, 2016

The flight deck duties are assigned as "Pilot Flying" and "Pilot Monitoring". Seems this approach eliminates pilot flying and consequence is lost of redundancy. Perhaps it is a matter of paying two pilots vs one so management can get bonus for cutting cost.

on Nov 18, 2016

Of course, Astro Droids!

on Nov 18, 2016

Back in the 1980's, DARPA sponsored the Pilot's Associate program that demonstrated all the fundamental architecture needs of an R2D2 for the back seat of a single seat fighter. Most of the hardware would be obsolete today, but the software architecture should still be valid.

on Nov 18, 2016

We already have autopilot. All the things that a robot could do are already done automatically by the systems. This is analogous to the old idea of a household robot, something to run the vacuum cleaner (we call that a Roomba now) and switch the oven on (most have timers) or wash the dishes (my Bosch dishwasher is almost silent and I have a teenager for packing it). Do you want a flailing mechanical robot in the copilot seat when something goes wrong? As an engineer I'd take exception to my glorious toy being characterized that way, but as an engineer who worked on safety critical systems for twelve years I also learned that I don't have perfect foresight.

on Nov 18, 2016

The primary lesson I learned on the Pilot's Associate program was the fallibility of human pilots. When you put them in a high-stress environment with too many "threats" to deal with, they get "tunnel vision." They pick what their instincts say is the worst threat and work that problem. However, a more rigorous situation assessment is necessary to ensure that all threats are considered. You don't need this capability very often in commercial flight, but when you do, it can save a lot of lives.

on Nov 18, 2016

But at what point does the AI run out of ideas? It the aircraft manufacturer says that a certain condition, or multiple-layered emergency cannot happen, is that where the algorithms stop, and the "could never happen" scenario goes beyond the ability of the automated co-pilot to think itself into the equation and actually be of any help whatsoever, or to at least offer an opinion in the decision making process?
Put AI into the right seat of QF32 and ask yourself where that scenario might have ended up. On that Qantas flight out of SIN when an engine blew and the damage and dynamics were beyond what Airbus believed could ever happen, the inadequate software driving the ECAM was unable to correctly assess and prioritise the multiple warnings as far as that combination of events affected the jet, and the limited scope of the FMC programming data refused to accept inputs. So much for that AI - if it had been driving an automated copilot, the thing in the right seat would have been trying to fix what the software thought was the most pressing problem while ignoring the really big issues further down the ECAM page. In the case of this incident, it took the combined skills and CRM inputs from the captain, a first officer, a second officer, and two senior check captions in the cockpit to sort out the mess, manually resolve a landing distance problem (that the aircraft software refused to compute automatically because it was unable to think through the problem) that indicated they would only stop 100m from the end of the runway, and finally save the aircraft.
AI in the right seat....? Risible! It might work when they build the perfect aircraft, but how far away is that?

on Nov 19, 2016

What about the two men rule for the cockpit? So at any time a flight attendant has to seat in the cockpit and sometimes even two. Wages are somehow different.

In case one automated pilot is quite OK. Why not two automated pilots and no human pilot as all? As long as two automated pilots are not better than two pilots the call for one automated pilot is just the greed of airline industry to save money.

I have no problem with a third automated pilot entering the cockpit but what could be the difference to the already existing autopilot with an additional external view? To read the instruments via a video interface is nonsense for an automated pilot. An automated pilot can read the data in a more direct way. What happens in case someone puts his jacket about the lens of the automated pilot? What will the automated pilot do? I can imagine the answer of a pilot with a jacket put over his eyes.

on Nov 19, 2016

I'm not a pilot. I'm thinking of an initial trial program with 1 or 2 robots in the cockpit for cargo only flights. No lives at risk. Robo-1 would be master pilot and Robo-2 could be backup. Robo-pilots could be monitored for errors over several years. If needed ,
software/hardware updates could be made during this trial period. Let's learn if the robo-pilots could be made error proof/100% reliable before humans are on board.

on Nov 20, 2016

I am a pilot only a glider and prop one but not a airline one but I prefer having an screwed up homo sapiens flyicus than a robot fully loaded of algorithms!!

on Nov 21, 2016

Machines have much lower error rates than humans, it is only a matter of time to when human pilots serve only as system managers and their real function is to demonstrate human skin in the game (most people would never board a flight where a human operator isn't present).

on Nov 25, 2016

The next generation 50 passenger regional jet only will be economical if single-pilot. North America would be a perfect place to implement such thing.

on Nov 30, 2016

Hmmm public acceptance, I think not.

on Dec 6, 2016

Over 50 years ago the ALPA magazine featured a cartoon showing a cabinet with a glass door and an airline pilot inside. A sign read "In Case of Emergency Break Glass."

Articles in Aviation Week over recent months reveal the near future. The many pointing towards autonomous flight. In particular the one featuring the robot copilot.

A development that will mean the services and airlines will not have to wait for a new generation of airplanes to take advantage of the "autonomous airplane."

Within a decade statistics will show "driverless" cars are safer and there will be demands to do away with human drivers.

The "autonomous" airliner cannot be far behind.

IBMs Watson has written a hit song. Artificial Intelligence is now quite capable of writing articles for Aviation Week.

What was speculative 25 years ago, is on the verge of becoming common.

The "Expert System" MYCIN of 1974 was limited by it's database and processing power. Still it was better able to diagnose and recommend treatments for bacterial blood disorders than doctors, within strict limitations.

Now improvements in AI have reached the point where programs can better screen patient data, recommend tests and analyze the results than doctors.

Robotics have eliminated many factory jobs and are on the verge of eliminating most remaining middle class jobs over then next few decades.

While I don't fear as Stephen Hawking suggests, that AI will end the human race; it is clear it will end the middle class as we know it.

It will even end many of the lower class jobs we know. The day is not far off when your driverless car will roll you up to the order window and the McRobo will ask "do you want fries with that"? That is while the previous customer taps their smart phone and the McRobo dispenses his order.

The savings in labor costs will enrich the few. That same savings will impoverish the many.

We have seen occupations go away before. The PBX operator, the elevator operator, service station attendants, these were common when I was a child.

At various times they were eliminated.

But history has never seen such a mass elimination of good paying jobs in such a short time as is imminent.

What will the consequences be? Will a former airline pilot be happy if he can find a minimum wage job?

What is the world going to do with all those unemployed citizens?

Will we be able to disenfranchise them all before they revolt?

Will General Ned Ludd rally an army and ride to the rescue?

We have seen what a demagogue may do with unemployment and resentment.

Please or Register to post comments.

Aviation Week & Space Technology
Penton Corporate

Sponsored Introduction Continue on to (or wait seconds) ×