In article ,
says...


Okay, let's accept that this system is built. What happens when
communication is interrupted? Radio failure is not an unheard of
event, is it?

In the case of communication failure, it would appear that you have
two choices. You could have the planes automatically go into a
holding pattern of some sort, or you could have the planes act
autonomously. If you went into a holding pattern, the planes would
have to be able to break out of the holding pattern and land
autonomously if they ran low on fuel or detected incoming weather.

Having a pilot on the ground remotely controlling the plane does not
remove the need for autonomous operation -- it just means that the
autonomous operation is only used during unusual situations. I
believe that handling these unusual situations are exactly parts of
the autonomous controller which will be the most difficult to design
correctly.

So, you are left with two choices: 1. Try to design a communication
system which is so robust that communication failure is virtually
impossible; or 2. Include some sort of autonomous system as a backup
for when communication fails.

Do I think this is impossible? No. Do I think it is quite hard to
get right? Yes. It certainly will take quite some time to get this
right enough to win the trust of the average passenger.



This I agree with completely, and it echoes my initial post to this thread
explaining why we would not desire to create such a system. Even if the
automated system delivers better safety performance overall than the
human-operated system, we will always be able to imagine a scenario where the
machine would not have an adequate response. Human operators create accidents
routinely, where no underlying emergency existed, and we accept this because,
well, we're all human. We would not be ready to accept, however, even a small
number of machine-induced accidents, in cases where a human operator on board
may have been able to save the situation.

We can easily imagine a response to the challenge you've just posted. But
where one is resolved, there are dozens of others waiting to surface. Making
the aircraft more "autonomous" is the wrong answer - it only makes the
equation more complex. Imagine if we've lost communication to the plane,
and can no longer tell it what to do. It's stuck in a holding pattern. Not
optimal. Now, in addition to this, imagine it is, to some degree, autonomous,
and is faced with a situation its designers didn't anticipate. Not only can we
no longer tell it what to do - we don't know what IT's going to do! That
sounds like real fun! Just one machine-induced disaster will have everyone
clamoring for a return of a pilot in the plane.

G Faris