"Pilotless airliners safer" - London Times article

[Tourist]

Just a quick one.

There seems to be an assumption that I "like" automation and unmanned aircraft.

I don't. I loath the things.

They suck the fun out of life.

I left airline flying because I enjoy flying and being a pilot. I now fly a real old school aircraft again that needs me.

No autopilot
No ECAM
No TCAS
No EGPWS
No CVR
No ACARS
No magenta line
No mode S
No weather radar


What I don't do is fool myself that I'm better than automation.

Thankfully, since there are not that many of us left that can fly without the toys I get paid very nicely to do it!

[Greenlights]

I think, the problem is that pilots are not objectives in their point of view.

Personnally I don't care much what will happen in the futur, but for sure, you bet, there will be pilotless planes.

first of all, pilots are not even engineers to give an opinion about what a technology can do or can not do...science can go very far, more than we know. Unless you are a scientist.

2) human being, in all cases, always like challenge. We always want to go further and create new technology. It's an exponential curve.
As many doubt about it , that pilotless planes are difficult to make, yes maybe. In that case, it means it's a challenge. So you can be sure that engineers want to win this challenge.
A challenge is what makes us motivated. Keep this in mind.
It seems impossible for you now. But in the futur it will not. Your kids will tell you that.
It is not honest to think that such things won't happen. As long as we have brain, then, we will still invent new technology, better than today.

[Tourist]

Ancient

Not sure who that is, but fail to see what relevance a single pilot getting something right has?
Plenty get it wrong.

the list of pilot error crashes that a computer would never have made is enormous

Asiana
Colgan
Shall I go on?

It makes no difference, to my argument. I conceded that there will be occasions, the black swan events where a human would be better.
Most accidents are not something new.

It's not that the best pilots are not great, it's that the worst are so much worse than a computer.
Computers are consistent. You can produce 1000 and they will perform the same day and night, and if they don't, you swap them out!

[kcockayne]

Tourist

I don't know you , or your piloting skills, but trust me, you are FAR better than a computer !

[Superpilot]

My final conclusion...

No, pilotless airliners are not safer

My reasoning is that absolute safety can only be guaranteed by intelligence, not by speed and efficiency or a "computer that never gets tired". Intelligence has the edge and no amount of AI will ever beat it because the AI we develop will contain the inherent limitations of our minds, at least in ability. If you have ever coded an application you will know what this means, if you haven't then keep banging on about your DLR and UAV examples till the cows come home.

An automated airliner carrying 100+ pax takes off (and lands) might be commissioned before I leave this planet but it will be on the basis of the accepted reality that it is "safe enough". By then, every aspects of our lives will be so automated and so "local" that air travel would be a thing of the past anyway.

* intelligence being that thing that computes multiple inputs and outputs to achieve the correct outcome in a previously undefined situation

[Tourist]

Thank you for your vote of confidence, but nope, there are plenty of things I can do that computers cannot currently do, but in the range of skills that the automation can currently do in aircraft, I cannot think of something I am better at.
Sometimes I land better than autoland, often worse, certainly less consistently.
I cannot fly an approach as well.
I burn more fuel
I cannot fly with unwavering attention for 17hrs.
Sometimes I need to get up to pee.

[Tourist]

Superpilot

Go read up on what an AI is.

I don't think it means what you think it means.
I wouldn't want an AI in the cockpit.

What are you talking about absolute safety for?
There is now way to have absolute safet.
We don't have it now and we never will.

[Superpilot]

Thank you but I know exactly what AI is. I just read the wiki article and it concurred with my thinking

Computing and software/hardware development is the other career I still run in parallel to my flying career, therefore I'm not just another steam guage era jet jocky with a mentality to defend.

You are right, absolute safety can never be achieved but safety itself is better achieved through employing intelligence (the natural kind being better than the artificial kind).

[Pininstauld]

oh dear

To the folks who see trolls behind every comment they disagree with, stop posting ad hominem and address the issues. FYI, I have +10k hours, am a qualified TRI on the A320 family and now teach conversion courses. The point I made about ECAM is valid, because most of the time ECAM has already closed a valve, opened a contactor, isolated a faulty component, etc. and the pilot is just changing a FAULT caption to an OFF caption when she follows the drill. No one is arguing that ECAM is perfect - that's why the QRH is still so thick - but that doesn't alter the original design goals: to use automation as much as possible to reduce workload and minimise human errors. We are on a journey and the end of our journey is fully automated passenger flights. Not saying I like it any more than anyone else. Just making obvious comments on a very sensitive issue. If that's trolling to some, well - just thanks for your honesty.

[RetiredF4]

It's not that the best pilots are not great, it's that the worst are so much worse than a computer.
Computers are consistent. You can produce 1000 and they will perform the same day and night, and if they don't, you swap them out!

Your examples are all single point references, which prove nothing. The single train line from A to B and back, the single car driving with limited speed, the military UAV's for special missions. Everybody can have a quadrocopter and program it to fly from A to B and back, carry a camera or a small box.

That does not equal to a complete pilotless system. What good is it when the computer runs without probs ( which no computer does), but the input devices and the technical stuff which makes a functioning airframe has not the same reliability? There is no pitot boom on the market which is fail safe and no engine. All sensors are prone to fail once now and then, and sometimes even at the same time.

To make such a purly computer driven aircraft safe, the redundancy of the systems would have to increase significantly. The amount of software and database to cope with all possible failure modes and all asociated variables would exceed any imagination. You mentioned the landing in the Hudson. Who would have foreseen that? Landing in the water, yes. But landing in that spot of water? No. To cover all these possibilities zillions of options would have to be taken care of, any strip of water on this earth suitable for such a landing would have to be foreseen and acounted for, under any weather phenomena and under all the different aircraft failure modes. Your statement that a computer would have done it better is as true as it might be is useless, because no programmer and no software engineer would have been prepared for such situation, so no software code written. And without that pre programmed database any computer system is just blind. You would need an artificial self adapting and self learning intelligence system to be prepared for all situations.

Computers are almost consistent, and all 1.000 of them produced may work identical. One software bug discovered in one of them (hopefullly not by a smoking hole) will ground all other 999 until new software is developped, written, tested and uploaded. That will be a bad day for shareholders.

How difficult, time consuming and cost intensive the developping and implementation process of software to a required standard (even with pilots as fallback position) turns out to be can be seen on the status of the F35and the Typhoon. The jets fly since years, the software development for them is still on its way.

The so-called Apollo Guidance Computer (AGC) used a real time operating system, which enabled astronauts to enter simple commands by typing in pairs of nouns and verbs, to control the spacecraft. It was more basic than the electronics in modern toasters that have computer controlled stop/start/defrost buttons. It had approximately 64Kbyte of memory and operated at 0.043MHz.

My first computer 1985 was a commodore PC10 III and it was far better than the one astronauts in the Apollo spacecraft had available. Todays computers are far better, smaller, faster and cheaper and the software can do magic things. But they are still just man made systems with man made software and man made failures, and if they should operate a passenger transporting aircraft around the globe on various routes, different times and in different weather conditions, then you need not only more reliable computing capacity, but the reliability and redundency of aircraft systems have to improve as well. As other industry changing inventions have shown the initial momentum of improvement slows dow, because some day it reaches the end of its economical development potential. To assume that the speed of further development continues to increases over time is misleading.

Manned aircraft achieved Mach 1 in 1947, Mach 2 in 1953, Mach 3 in 1956, Mach 4 in 1961. In the year 1970 the first supersonic comercial airliner Concorde flew Mach 2. It is amazing that 45 years later we have no supersonic transport aircraft in service despite all those technical advances and this vast computing power available.

Would we like to see even % of that loss rate of present UAV's?

http://dronewars.net/drone-crash-database/

[16024]

Now we are getting close to the point.

[Tourist]

Thank you for at least bothering to research and link.

A question.

If they decided they were losing too many reaper/predator and built a cockpit on top and sat a pilot in it.

What do you think he could do to prevent a crash post engine failure in his aircraft?

They are single engine aircraft with very limited/zero redundancy operating in a warzone.

However, I suspect that if you were to remove crashes where the failure made the type of pilot irrelevant, the picture would still make their loss rate worse than airliners.

Almost as if......
They are built to a lower safety standard because they don't carry people?

If you built a passenger carrying version the drivers just might be a bit different?

I would not use PPL aircraft crash figures to show how dangerous airline flying is because it would be disingenuous.

Incidentally, early piloted aircraft fell like flies too.
They got better though and then passengers got on.

Plus of course these are not autonomous. They are remote piloted. Totally different challenges.

[RCav8or]

I wonder how a computer flown A/C would have handled the situation an L-1011 pilot found himself in departing LAX, back in the early days of the L-1011 service life. The bearings for the full flying stab froze in the position called for at rotation. The pilot avoided the stall they were headed for by pulling back power on the wing mounted engines and increasing thrust on the rear engine. They managed to get the plane on the ground in one piece by learning to control pitch by changing the relative thrust between the rear and wing mounted engines.
A pilot of that caliber, combined with the best best in modern automation is how I want to get from point A to point B as PAX.
Pete

[RetiredF4]

@ Tourist

Those who believe the technological challenges are insurmountable should ponder on this.

Various militaries in the world are testing autonomous UCAVs
These are unmanned, autonomous combat aircraft.

Autonomous X-47B Flies In Formation With Fighter Aircraft | Popular Science

These aircraft must find must take-off (from a carrier in this case) navigate, find and kill the enemy, navigate home and then land on.
No beacons, no radio control, no atc, no TCAS.

You should have thought about that before you posted the above, your optimistic reference to the military use of unmanned vehicles brought me into this discussion. If you want to go into detail to what reasons those UAv's had been lost, go ahead. For sure software and computers didn't make them safer, it only saves the sad duty to visit one more widow.
As I mentioned before, these drones the military is flying are not existent because the pilots are bad and the computers are better.

[BBK]

Visions of the future

Does anyone remember seeing programmes like Tomorrow's World and the like? Back in the seventies we imagined we'd go to work in a suspended monorail module, there would be regular flights to a base on the moon, supersonic flight would be routine and we would all be served at home by a humanoid robot like C3PO!

As I said before "never say never" but predictions about the future were often horrendously inaccurate. Of course technology has advanced but often where we didn't envisage it. I used to work on a mainframe computer that had a storage unit that was the size of a large bookcase. It had a ferrite core and 512k of memory. The processing power, storage capabilities and the Internet have revolutionised our way of living. However, in terms of autonomous systems I imagine that we will see UAVs being allowed into mainstream airports in a "mixed mode" and I suspect that is where a lot of the research is going. However, the sheer complexity of decision making on a typical flight to say a busy airport where the presence of storms means traffic is being vectored all over the place and then diverting etc makes for an incredibly complex situation to model.

I imagine the autonomous systems would rely on a rule set, devised by a possibly fallible human, but we (humans) use judgement, experience, knowledge and we can learn. As someone said before any aircraft will HAVE to be remotely accessible eg in the case of a reroute so fully autonomous (probably) isn't feasible. I understand why the military value UAVs but there are far more questions than answers as to the desirability of such aircraft. I hope the aircraft designers will do all they can to put humans back into loop not take them out of it.

[Tourist]

RCav8or

And I wonder whether anybody bothers to read previous posts before jumping in.

Post 113 covers it at length. That sort of thing is what computers are good at and systems to deal with it are not flying operationally. Essentially almost invisible to the pilots.

RetiredF4
I will say it more clearly.
The military is flying autonomous UAVs
The ones that make up that list as far as I can see are not the autonomous ones.
Do you see X-47B on that list?

None of this is relevant because nobody has designed any of them with the kind of safety and redundancy you would put in a passenger UAV.

[Tourist]

BBK, all fair points

I do however remember they had automatic windscreen wipers on tomorrow's world that never worked in the studio.

They did eventually hit the mainstream though....

[MG23]

Originally Posted by Tourist

Computers are consistent. You can produce 1000 and they will perform the same day and night, and if they don't, you swap them out!

You say that as though it's a good thing. If they fly the same way, they also crash the same way, if there's a rarely-triggered bug that got through the testing.

[Tourist]

Fortunately, rarely triggered bugs are rarely triggered, so you have time to patch the software before the second smoking hole.
Doesn't stop the first hole of course, but then you can't have everything.

When humans make a smoking hole, history proves that another one and another one will follow until they invent something to stop it.

EGPWS for example!

An entirely automated system that checks if the pilot is being a prat and tells him what to do to not die.

Sometimes, because it cannot actually take control off the pilot, they still manage to hit the ground, but mostly when they act as controlling monkey for the fully automated system they don't crash.
It's very effective and one of the reasons CFIT has dropped spectacularly.

TCAS

Another example.
It tells the pilot what to do. It requires no thought, in fact it specifically says not to think, just do what you are told and you will not hit anything.

Again, it would be easy to integrate with the autopilot and then it would be even more successful but pilots might complain.

Do you notice a trend?
We make an automated system that requires no human thought and things get safer?

[slast]

Where's the business case?

Basic reason this is not actually going to happen in the lifetime of any current PPruner is: No Business Case for the foreseeable future.

It will happen when customer demand can be met at a sufficiently low price for manufacturers to make money supplying and operators to make money using them. That will depend on price and demand.

However fancy the tin brain, the price will depend heavily on certification cost which will probably be very very high.

As for demand,

(1) Are airline customers (pax) demanding it? No. I don't see demonstrators outside even Ryanair's office with placards saying "£5.99's too much, get rid of those pilots".

(2) Are airline customers (freight) demanding it? No but they can see more possibilities because (1) doesn't concern them.

(3) Are airlines demanding it? No but they will monitor any developments so as not be caught out if there is a change in (1) or (2)

(4) with no immediate customer demand, manufacturers are hedging their bets by putting small amounts into watching developments by non-airline players.

Why should we care about ill-informed journalists pontificating in the press?
Because it is distracts attention from fixing things that really need fixing, like getting instrument vertical guidance on every approach used for commercial aviation.

A lot of the discussion has focussed on landings. What about takeoff?

Autopilot takeoff is trivial compared to autoland. I first took part in autolands with pax on board 50 years ago, but 3 aircraft generations later no-one has built a commercial aircraft with an auto takeoff system - there's no business benefit to doing it.

Lifting max t/o weight is the most important parameter in the business case for buying a specific aircraft. Certification is based on engine failure only and does not take full account of many other factors (variable surface conditions etc). Basic performance standards were written in the 1960s (ICAO SCAP /Airworthiness Committee etc) and no commercial enterprise (manufacturer or operator) wants to reopen that issue. That's why grandfather rights and "new models" of old types are so important. However any auto system implementation will have to recognise and find a way to deal with these. That will inevitably result in lower max weights off the same runways - probably for ALL commercial aircraft not just pilotless. (if Hence no business incentive to develop auto takeoff.

Question for the YES side.

A pilotless airliner with 400 pax. is taking off in the correct configuration. At VR a system detects a changed status of a physical system. A/C protection software automatically reconfigures aircraft to settings appropriate to changed status. The new configuration puts a/c immediately outside the flight envelope and makes it almost incapable of flight. This condition had not been considered in any certification fault tree analysis.

(a) what do you envisage would happen to this pilotless a/c?
(b) which organisations will volunteer to be liable for the consequences?