Unmanned cargo flights Was: Prognosis Weak for Tonight's MarsLanders

Henry Spencer wrote:
In article ,
Dick Morris wrote:

Unmanned cargo rockets do not need to be expendable, nor do they need
wings to be recoverable. Vertical landing with rocket braking is a
perfectly reasonable alternative.


Once the vehicle becomes recoverable, its value goes up
considerably, and
a reasonable designer starts considering putting a pilot on it, just
because he might sometimes prevent it from being lost. (How many
unmanned
cargo aircraft do you see?)

Good question.
I assume there are quite some cargo aircraft that have
1) auto-pilot
2) fly by wire
3) continuous radio contact to flight control
4) advanced sensors like collision radar, GPS, height to ground radar
5) receivers for radio beacons to land blindly

So why are there no unmanned cargo planes?
My guess:
a) cheap life support systems
b) tradition (legal reasons, unions, insurances)
c) take off/flying/landing in difficult areas
d) missing infrastructure for reliable remote control in emergencies

Wasn't Buran landed remote controlled the first flight?

Autonomous and tele-operated robotics have progressed a lot since the
space shuttle was built. From my point of view, a semi-autonomous
robotic space craft looks pretty reasonable.

In both shuttle disasters, the human pilots didn't make much of a
difference. Of course I don't know how many critical situations in the
shuttle were successfully controlled by the crew. Does anyone know?

But how many of them couldn't have been managed by ground control?
(Given the shuttle would be equipped for that.) You usually can't do
any repairs in critical flight situations, can you? You look at the
instruments and decide which key to press and which way to move the
joystick.

If a new space craft is designed for fly by wire, why not design it
for remote control? If it carries a manned cargo, you can plug in a
control set for the pilot. If it does an unmanned flight, let it fly
by autopilot, which does all decisions that need reactions within less
than one or two seconds. Decisions like to abort a launch, switch to a
backup system, etc. can be radio controlled from the ground.

A medium or even high bandwidth radio connection to a space ship can't
be that expensive these days. Of course there are some minutes during
re-entry without radio. But I'd say that this can be handled by
auto-pilot pretty well. It's a dangerous phase, but not a difficult
one, from the point of view of the auto-pilot. The aerobrake works or
it does not. No way to abort.

Pilots of airplanes are especially useful for choosing emergency
landing spots. But with vertical landing, you should have a good view
pretty early, and requirements for the landing spot are much easier to
meet. Transfer an image from the downward camera, click a position and
the auto pilot will try to go there. Repeat this every five seconds or so.

Fully autonomous emergency landing would be a admittedly more risky. I
attended a talk about a laser range scanner that is used for
helicopters. It could be used to look for a flat landing place.

But on the other hand, if your flight trajectory can't be aborted to a
designated emergency landing place with radio beacons _and_ radio
contact is lost, something did went seriously wrong. How likely is
such a situation and how likely can a pilot save the spacecraft in
such a situation? Space shuttles have much less options where to fly
than aircrafts have.

I would say unmanned flight is more a psychological problem than a
real decrease in safety.

In article ,
Axel Walthelm wrote:
a reasonable designer starts considering putting a pilot on it, just
because he might sometimes prevent it from being lost. (How many
unmanned cargo aircraft do you see?)

Good question.
I assume there are quite some cargo aircraft that have
1) auto-pilot
2) fly by wire
3) continuous radio contact to flight control
4) advanced sensors like collision radar, GPS, height to ground radar
5) receivers for radio beacons to land blindly

Fly by wire is not that common on airliners, and most cargo aircraft are
airliner conversions. Nor is any great point made of maintaining
*continuous* contact with the ground; "flight control" is the pilot. But
skipping those and some other details, yes, basic technical feasibility is
clearly within easy reach, when everything's working.

So why are there no unmanned cargo planes?
My guess:
a) cheap life support systems

Not necessarily cheap in an absolute sense, but relatively cheap because
of the use of airliner conversions (the development is already done).

b) tradition (legal reasons, unions, insurances)

Much of this is not so much tradition, as safety when things go wrong.
Laws and insurance rates are responses (albeit slow ones) to real risks.

c) take off/flying/landing in difficult areas

Not, in itself, anything that's terribly difficult for automation.

d) missing infrastructure for reliable remote control in emergencies

Reliability is the big snag with remote control. It's not so much a
matter of missing infrastructure, as of inadequate fault tolerance. Only
a few years ago, the Perseus B drone aircraft, being flown from Dryden,
crashed on an interstate highway near a town (luckily, nobody was hurt)
when both its flight controls and its flight-termination system failed.
This sort of example is not encouraging to the regulators and the
insurance people.

There is also a lesser issue of possibly-inadequate situational awareness,
given the limitations of the data being sent back. (E.g., noise and
vibration are significant information in the event of engine trouble.)

The question is always: is this remote-control system so good that there
is *no* advantage to having a pilot on board? So far, the answer to that
has always been "no way". The mere possibility that the pilot might save
a hundred-million-dollar aircraft is ample reason to have him on board.
(Losing a pilot is much cheaper than losing an aircraft.)

Wasn't Buran landed remote controlled the first flight?

No, it used automatic landing. (Which was first tested very extensively
in manned low-altitude flights, using the jet-powered Buran trainer.)

Autonomous and tele-operated robotics have progressed a lot since the
space shuttle was built. From my point of view, a semi-autonomous
robotic space craft looks pretty reasonable.

If it's reliable enough to carry billion-dollar cargo, it is reliable
enough to carry a pilot. And if it's reliable enough to carry him, then
having him there makes all kinds of sense.

In both shuttle disasters, the human pilots didn't make much of a
difference.

There are always unsurvivable failure modes. Pilots have to have useful
options for them to make a difference. When the mess is bad enough, there
may not be any.

Of course I don't know how many critical situations in the
shuttle were successfully controlled by the crew...

It's hard to count, because such things don't necessarily make the news.

But how many of them couldn't have been managed by ground control?
(Given the shuttle would be equipped for that.) You usually can't do
any repairs in critical flight situations, can you? You look at the
instruments and decide which key to press and which way to move the
joystick.

The joystick usually isn't the big issue. It's the switch-flipping --
deciding which systems, and which modes, have the best chance of dealing
with the situation. For example (a real shuttle example), overriding
engine temperature sensors, so they don't cause an engine shutdown,
because it's clear that they are mistaken.

And yes, this can *IN PRINCIPLE* be done from the ground. But there is
currently a very large gap between principle and practice here, and you
can't close it just by waving a magic wand and commanding "begone!". When
things go wrong, communications are often one of the first things to fail.

If a new space craft is designed for fly by wire, why not design it
for remote control?

For the same reason that it's not done for aircraft: reliability. Having
the pilot on board, rather than on the ground, is a big reliability win
with today's technology and likely near-future technology.
--
MOST launched 30 June; science observations running | Henry Spencer
since Oct; first surprises seen; papers pending. |

(Henry Spencer) writes:
The question is always: is this remote-control system so good that there
is *no* advantage to having a pilot on board? So far, the answer to that
has always been "no way". The mere possibility that the pilot might save
a hundred-million-dollar aircraft is ample reason to have him on board.
(Losing a pilot is much cheaper than losing an aircraft.)

You're going to start the manned versus unmanned space shuttle thread
again. ;-)

Seriously though, this same argument can be made with the shuttle.
Losing one of your three remaining (essentially irreplaceable)
shuttles is worse than possibly losing another crew. The crew
enhances the safety of the shuttle for the reasons you mention.

Jeff
--
Remove "no" and "spam" from email address to reply.
If it says "This is not spam!", it's surely a lie.

Henry Spencer wrote:
a reasonable designer starts considering putting a pilot on it,
just
because he might sometimes prevent it from being lost. (How many
unmanned cargo aircraft do you see?)


c) take off/flying/landing in difficult area


Not, in itself, anything that's terribly difficult for automation.


Given sufficient guidance from the ground, yes, easy.
For a VTVL space shuttle it should be ok.

But airplane pilots prefer to look where they go. This adds safety and
currently can not be replaced by computers.

Compared to a space shuttle, airplanes travel much longer close to the
ground, within weather and close to other airplanes.

d) missing infrastructure for reliable remote control in emergencies


Reliability is the big snag with remote control. It's not so much a
matter of missing infrastructure, as of inadequate fault tolerance. Only
a few years ago, the Perseus B drone aircraft, being flown from Dryden,
crashed on an interstate highway near a town (luckily, nobody was hurt)
when both its flight controls and its flight-termination system failed.
This sort of example is not encouraging to the regulators and the
insurance people.

I guess we can find examples for accidents that happened because
a) the auto-pilot messed up
b) the human pilot made a fatal mistake and
c) the communication between auto-pilot and pilot went wrong.

The difficult question is the probability of each case.

There is also a lesser issue of possibly-inadequate situational awareness,
given the limitations of the data being sent back. (E.g., noise and
vibration are significant information in the event of engine trouble.)


Good point. "Flying with your ass" is much more difficult remote.
You save weight for pilot, life support system and cockpit. But you
should add some more sensors and better radio communication equipment.
Overall you will still save quite some weight, and you don't have to
man-rate the flight, so you can save even more. As long as each kg
saved is worth thousands of dollars ...

Intuitive flying is also a matter of training. You could (and probably
should) connect the remote data to a full fledged flight simulator,
including vibrations and derivates of acceleration. Gyroscopes and
microphones are well developed sensors. Sensors for smell are not as
good. Filling the simulator with smoke as soon as smoke is detected
on-board would probably be kind of overkill :*)

There is even added security by not exposing the remote-pilot to
accelerations and vibrations that hinder him from doing the right
thing. No more risks of life support failures. A hull breach would not
hinder the pilot either.

The question is always: is this remote-control system so good that there
is *no* advantage to having a pilot on board?

Especially for cargo space flights the question should be: is the
added cost of an onboard pilot larger than the average loss due to not
having an on-board pilot.
Where is the break even point between spending more money on safety
and saving money by having a re-usable shuttle?

I guess this point is pretty much on the "safe" side. But not at
"infinitely safe".

If it's reliable enough to carry billion-dollar cargo, it is
reliable

A billion dollar? This surely does include development costs for the
satellite or whatever you launch? Rebuilding it once more should be
much cheaper. Its still many millions of dollars, but much of this is
because launching is so expensive that it justifies as expensive
satellite systems. If launching could be made cheaper (which I assume
to be the purpose of building reusable space shuttles), freight
becomes cheaper too. Going for 99.99% safety is extremely expensive.
Finally 99% safety and a good insurance could very well turn out to be
cheaper (at least for cargo launches).


Wasn't Buran landed remote controlled the first flight?


No, it used automatic landing. (Which was first tested very extensively
in manned low-altitude flights, using the jet-powered Buran trainer.)

Programming a good auto-pilot is much more difficult than training a
pilot. But auto-pilot and its programming is much easier to copy.
Human pilots use flight simulators. And so should developers of
auto-pilots.

If you have an auto-pilot which almost always can handle the
situation, and you almost always have a working remote control to
overrule or correct the auto-pilots decisions ... is the remaining
risk that could be handled only by an on-board pilot really worth it?

with the situation. For example (a real shuttle example), overriding
engine temperature sensors, so they don't cause an engine shutdown,
because it's clear that they are mistaken.

First time this happens your chances are good that remote control
works.

Then it's time to go back to add more capabilities to the auto-pilot.
Situations like that can be classified by computers too, using
redundant sensors and models relating those sensors.
(Or maybe analysis will show that the risk of a double failure is not
worth the expense.)

And yes, this can *IN PRINCIPLE* be done from the ground. But there is
currently a very large gap between principle and practice here, and you
can't close it just by waving a magic wand and commanding "begone!". Whe
things go wrong, communications are often one of the first things to fail.

If this is so, it is because communication is not considered safety
critical enough to install more redundant and independent backup
systems.

Since most launches to space are currently done without onboard pilot,
I wouldn't say the gap is very large. But it surely needs some proper
engineering.

Magic wand waving is by no means my recommended approach to the
problem ;*)

Let me say that I am in favour of manned space exploration. But men in
space need a lot of equipment, and I hope that robotics can be used to
make manned space exploration more affordable.

Robotics can be applied in much less cases than common people would
expect. But in this case I would say it's reasonable to try.

Remote control from the ground of spacecraft during docking operations
is non-trivial, even if all systems are working perfectly. Something
about piloting with a time lag makes it very difficult.

(Henry Spencer) wrote in message ...
Autonomous and tele-operated robotics have progressed a lot since the
space shuttle was built. From my point of view, a semi-autonomous
robotic space craft looks pretty reasonable.

If it's reliable enough to carry billion-dollar cargo, it is reliable
enough to carry a pilot. And if it's reliable enough to carry him, then
having him there makes all kinds of sense.

Unlike many here, I favor unmanned vehicles for flight testing. To be
crass, it comes down to putting a value on the life of your pilot. If
your vehicle and / or cargo are worth a billion dollars, then adding a
pilot isn't going to modify your risk all that much, and most likely
will help things out. If your vehicle only costs $100,000, then
putting a pilot in will completely dominate your risk, and IMHO, isn't
a good idea. I would argue that holds for a million dollar vehicle as
well, but it gets fuzzier as you add additional orders of magnitude.

I think that most of the interesting work can happen with vehicles
that cost under $1 million (not program cost, just test vehicle
replacement cost), so I expect to be working on vehicles that are
tested remotely / autonomously for the forseeable future. They are
certainly intended to carry people, but not for the test flights.

John Carmack
www.armadilloaerospace.com

(John Carmack) wrote in message . com...

If your vehicle only costs $100,000, then
putting a pilot in will completely dominate your risk, and IMHO, isn't
a good idea.

The insurance industry seems to disagree, and they have a lot of data
points on $100,000 vehicles.

"Edward Wright" wrote in message
m...
(John Carmack) wrote in message
. com...

If your vehicle only costs $100,000, then
putting a pilot in will completely dominate your risk, and
IMHO, isn't a good idea.

The insurance industry seems to disagree, and they have a lot of
data points on $100,000 vehicles.

Could you elaborate? I would think that expensive car design is
primarily based around driver/passenger safety, not vehicle
preservation. I would expect Armadillo to follow a similar path, hence
extensive unmanned testing initially.

Pete.

Edward Wright wrote:
(John Carmack) wrote in message . com...

If your vehicle only costs $100,000, then
putting a pilot in will completely dominate your risk, and IMHO, isn't
a good idea.

The insurance industry seems to disagree, and they have a lot of data
points on $100,000 vehicles.

And they (rightly) even disagree even though the insurance policy on the
life of the pilot is likely to be in excess of $100,000.

--
Sander

+++ Out of cheese error +++

On Thu, 15 Jan 2004 19:01:19 GMT, (Henry Spencer)
wrote:


Reliability is the big snag with remote control. It's not so much a
matter of missing infrastructure, as of inadequate fault tolerance. Only
a few years ago, the Perseus B drone aircraft, being flown from Dryden,
crashed on an interstate highway near a town (luckily, nobody was hurt)
when both its flight controls and its flight-termination system failed.

By the highway, not on it. You make it sound as if it came down on
the roadway, but it was actually off to the side, in the desert. This
is just a demonstration of the "big desert" theorem, a cousin of the
"big sky" theorem.

Of all the crashes we've had at and around Edwards since I came to the
Antelope Valley, by both NASA and USAF, the only one I know of that
came very close to any sort of occupied structure was the X-31. And
even then, it was only a piece of debris that bounced up against a
garage wall without damaging it. That's a pretty good record
considering everything we've dropped in the last 45 years, including
the B-1, SRV, DAST, X-31, F-4s, A-37, T-37, F-18s, F-104s, T-38s, F-16
subscale models, ALCMs, and so on.

Mary

--
Mary Shafer Retired aerospace research engineer