By 2030, commercial passengers will routinely fly in pilotlessplanes.

Recently, Peter Duniho posted:

"Neil Gould" wrote:
In the same vein, piloted airliners are "good enough". The number of
catastrophic losses are quite small in comparison to the number of
flights. There is no evidence that aircraft piloted by computer
would fare any better, much less signficantly better.

What would you consider "evidence"?

I meant "evidence" in a loose way, not as legally valid terminology. ;-)
So, any vehicle capable of operating autonomously over long distances and
time could provide some "evidence", one way or the other.

There's no question automation would avoid certain kinds of losses;
the valid question (without an answer for the moment) is whether
human pilots balance that out with actions that a computerized pilot
could not take.

I am sure the pilots' unions will invest great resources in showing
that human pilots are better. But I'd just as soon see an
independent source for that analysis.

I see it a little differently. The contest is not between humans and
computer control a computer can fly an airplane autonomously from point A
to B. That's a ways off, considering the current state of AI.

As far as "good enough" goes, that's a social issue. For the time
being, I'd agree things are "good enough", especially the distrust
that the public would have with an fully automated airliner.

I also don't see this as an issue of public trust, because the mindset
that we have about such things today is not relevant. By the time AI has
achieved the required sophistication to pull this off, I'd expect that
autonomous machines would be quite the norm and everyone would be able to
accept the introduction of autonomous airlines as the next logical step,
pun intended. ;-)

But
long-term, airlines are looking at two things, at least:

* Overall loss rate
* Cost of operations

I am skeptical that the overall loss rate would change much, for the
reasons I stated in my last post.

As for cost of operations, it seems to me that support for autonomous
aircraft would require an even larger and more costly infrastructure than
the airlines have now. Who is going to service and pre-flight these
systems? Considering the number of service stations capable of dealing
with the problems identified by the computers in our cars and the expense
of repair, I don't think the airlines can expect to save much (if
anything) by eliminating pilots.

Regards,

Neil

In article ,
says...


I see it a little differently. The contest is not between humans and
computer control a computer can fly an airplane autonomously from point A
to B. That's a ways off, considering the current state of AI.



What AI?? The scenario you've just described is thousands of times simpler
than what happens every time you turn on your computer to check the
aviation ng. Taxiing is the only element in this scenario that is not
already fully automated, and performed better by machines than by people.
We only fly today to keep ourselves in practice, in case we "really" have
to fly "someday".

The question of pilotless aircraft is not a technical question. The
technical part is trivial. The question is when, and especially WHY would
we want to do such a thing? The revenue value of an airline flight makes
the cost of at least one pilot very small. Passengers will not want to fly
in an aircraft that does not have the ability to be taken over "in
extremis" by a human pilot, and insurance companies will never go for a
scenario where there is no hopeless sucker (especially a dead one) to take
the blame when something goes wrong.

G Faris

wrote in message

As far as high jacking: I think that a pilotless plane would be more
secure. The designers could put in a code or something and make it
impossible for a hi-jacker to take control of the plane.

Of course he could still blow the damn thing up in mid-flight, but he
wouldn't be able to fly it into the WTC or such.

Never again will a terrorist be allowed to fly a plane into a high value
target. The 4th plane on 9-11 proved that.
--
Jim in NC

Recently, Greg Farris posted:

says...


I see it a little differently. The contest is not between humans and
computer control a computer can fly an airplane autonomously from
point A to B. That's a ways off, considering the current state of AI.


What AI?? The scenario you've just described is thousands of times
simpler than what happens every time you turn on your computer to
check the aviation ng. Taxiing is the only element in this scenario
that is not already fully automated, and performed better by machines
than by people. We only fly today to keep ourselves in practice, in
case we "really" have to fly "someday".

I suspect that you are confusing "autonomous" with "automatic". There is
no question but that machinery can follow programmed instructions
precisely (that is at the heart of CNC), however that machinery is not
making decisions in a greatly dynamic environment. In your example, the
location of airports are fixed, and it is a relatively simple task to have
a set of instructions that would get an aircraft from one to the other;
OTOH, taxiing is a dynamic environment, requiring informational
interaction and control based on mutually agreed decisions -- i.e.
autonomy. AFAIK, today's systems are incapable of that. Weather is also a
dynamic environment, one which every aircraft must contend with on every
flight; course deviations based on developing weather also require
autonomy. How do current-day automatic systems handle that? AFAIK, they
can't.

One point of the DARPA challenge (cited earlier in this thread) is to
create autonomous vehicles capable of simply getting from point A to B in
a dynamic environment. The results speak for themselves.

Regards,

Neil

In article ,
says...


Recently, Greg Farris posted:



One point of the DARPA challenge (cited earlier in this thread) is to
create autonomous vehicles capable of simply getting from point A to B in
a dynamic environment. The results speak for themselves.


They do - and what they tell us is that navigating over uncharted terrain,
full of obstacles, is a challenge for land vehicles. Quite a different
challenge from taxiing across a few hundred feet of perfectly charted,
smooth pavement, custom designed to fit the particularities of your
vehicle!

If we really wanted pilotless airliners (and my argument is that we do not,
and will not, probably ever) it will be a simple matter to eliminate the
human-controlled taxiing phase. Simply towing the aircraft would be one way
to do it. Besides, ground operations are one of the most error-prone phases
of aircraft operations, usually because pilots misunderstand
instructions. So this would be a good candidate for automation, even
without getting rid of the pilots.

G Faris

As was pointed out earlier, I think the issue is more social then technical.
We have UAVs that are towed to a takeoff point, a computer program lauches
and later lands the vehicle and a pilot sees what the plane sees via
satilite. The flight is carried out, 99% by the program put into the UAV
before departure and the "pilot" takes over only if they see something on
the ground that makes them want to take a closer look.

The military has already started to launch hellfire missiles from UAVs
flying in Afganistan on orders from someone looking at a TV in Washington.

Before 9/11 I got to sit in the cockpit of a Dash8 from TO to YOW. In
response to changes from the tower, the pilot just entered the change into
the autopilot and it did the rest. The pilot did not take the controls until
he was about 500 AGL on approach. If the controller could just click on the
aircraft on his screen, type in the new heading or alt. and the computer
would do it that might simplify everyone's job with fewer communications
errors. You might still get a few people killed getting the bugs out of the
system, like when they sent the lander to Mars with one subroutine working
with KM/min and the other subroutine working in miles/hour.

Mind you at that point the highjacker just has to capture a control tower
and send a bunch of aircraft into the same location and doesn't get to be
with a single virgin, let alone 70.

So we should be clear on what constitutes autonomous, what parts would the
plane control and where would the pilot be who can override the computer.

The problem is that society will not accept a pilotless plane because of the
what if factor. What if the computer failed, who will reboot it, what if the
weather goes bad, etc. And the old quetion seen in all airplane disaster
movies, "who is flying the plane"?


"Greg Farris" wrote in message
...
In article ,
says...


Recently, Greg Farris posted:



One point of the DARPA challenge (cited earlier in this thread) is to
create autonomous vehicles capable of simply getting from point A to B in
a dynamic environment. The results speak for themselves.


They do - and what they tell us is that navigating over uncharted terrain,
full of obstacles, is a challenge for land vehicles. Quite a different
challenge from taxiing across a few hundred feet of perfectly charted,
smooth pavement, custom designed to fit the particularities of your
vehicle!

If we really wanted pilotless airliners (and my argument is that we do
not,
and will not, probably ever) it will be a simple matter to eliminate the
human-controlled taxiing phase. Simply towing the aircraft would be one
way
to do it. Besides, ground operations are one of the most error-prone
phases
of aircraft operations, usually because pilots misunderstand
instructions. So this would be a good candidate for automation, even
without getting rid of the pilots.

G Faris

Recently, Greg Farris posted:

says...

One point of the DARPA challenge (cited earlier in this thread) is to
create autonomous vehicles capable of simply getting from point A to
B in a dynamic environment. The results speak for themselves.


They do - and what they tell us is that navigating over uncharted
terrain, full of obstacles, is a challenge for land vehicles.

Apparently, you missed the part where most of the vehicles also failed the
"charted" course. In other words, they failed because they were not
capable of functioning autonomously. Regardless of the notions, acting
autonomously in a 2D dynamic environment will always be easier than
operating in a 3D dynamic environment. It really is a problem based in the
limitations of AI; we aren't there yet.

Neil

Recently, Andre posted:

As was pointed out earlier, I think the issue is more social then
technical. We have UAVs that are towed to a takeoff point, a computer
program lauches and later lands the vehicle and a pilot sees what the
plane sees via satilite. The flight is carried out, 99% by the
program put into the UAV before departure and the "pilot" takes over
only if they see something on the ground that makes them want to take
a closer look.

(rest snipped for brevity)

All that has happened is that the pilot is outside the cockpit. The UAVs
are *not* acting autonomously, which is a pre-requisite for _pilotless_
operation.

Neil

In article ,
says...



Apparently, you missed the part where most of the vehicles also failed the
"charted" course. In other words, they failed because they were not
capable of functioning autonomously. Regardless of the notions, acting
autonomously in a 2D dynamic environment will always be easier than
operating in a 3D dynamic environment.

Euh - we were talking about taxiing - on the ground.
Now, imagine you have full control of all environmental variables. No rules
- you can do anything you want. Plough it flat, remove all obstacles, pave
it, put down wires,tapes, optical encoders every two inches - you can even
gouge a groove in it and put a pin in the nose of your vehichle if you
want.

Do you still maintain that the present state of our AI is not up to the
challenge?

G Faris

Recently, Greg Farris posted:

In article ,
says...



Apparently, you missed the part where most of the vehicles also
failed the "charted" course. In other words, they failed because
they were not capable of functioning autonomously. Regardless of the
notions, acting autonomously in a 2D dynamic environment will always
be easier than operating in a 3D dynamic environment.

Euh - we were talking about taxiing - on the ground.

I was talking about autonomous operation. It doesn't matter where that is.

Now, imagine you have full control of all environmental variables.

No point, because that's never the case.

No
rules - you can do anything you want. Plough it flat, remove all
obstacles, pave it, put down wires,tapes, optical encoders every two
inches - you can even gouge a groove in it and put a pin in the nose
of your vehichle if you want.

Do you still maintain that the present state of our AI is not up to
the challenge?

Look at the results of the DARPA challenge, and decide for yourself.

Neil