Re: Driverless Car ?!

Yevgen Barsukov wrote:
On Dec 18, 6:47 am, jmfbahciv <jmfbahciv@aol> wrote:
Yevgen Barsukov wrote:
On Dec 17, 6:19 am, jmfbahciv <jmfbahciv@aol> wrote:
Yevgen Barsukov wrote:
On Dec 15, 6:12 pm, Bernd Felsche <ber...@xxxxxxxxxxxxxxxxxxxxxxx>
wrote:
j...@xxxxxxxxxxxxxxxxxxxx wrote:
Ian Parker <ianpark...@xxxxxxxxx> wrote:
As far as "when will they be on sale?" To some extent that
depends on legal/political considerations as well as technical.
At the present state of technology you can have automatic freeway
lanes, their introduction would involve a political decision. One
thing is clear, a road should if possible be completly automatic,
automatic and non automatic driving should not mix. This means
that after a set date all cars will have tio be equiped with a
certified system.
Not "should", "must"; mixing is a receipe for disaster.
There are legal questions about the libability for accidents. These
would generally be fewer than now, but there will still be some
failures. A legal framework needs to be developed.
The legal framework is already there.
Assuming that there would be fewer crashes with automated cars on
public roads is based on what? That the automated systems will make
fewer mistakes?
They are unable to willfully break the rules.
<snip>
It's code and code breaks very easily because the physical
constraints change.
Physical laws involved in driving are pretty much just
Newtonian physical that have been around for over 200 years. They
don't change and can be hard-coded.
Measured parameters that are put into the Newtonian
formulas do change, but that will not break the code,
it will be used by the code.
Non-physical issues such as negotiating between several
cars about who has priority, overtaking etc have to
be programmed and this can break (and will be improved
over and over). But humans are also not
infallible in making such decisions. At least robo-car
will not drive through a 4-way stop sign because it does not feel like
stopping today. It will get into negotiation, and it is much less
likely to fail because it does not involve somebody's mood, how
insecure somebody feels this morning etc.
Sigh! You are not thinking at all. People are still writing
Y2K bugs. That involves dates. Do you really think that people
won't write bugs involved time? And what about the coders who
have absolutely no idea that there is a large difference between
the length of a meter and the length of a yard. Haven't you been
paying attention to what is written in these newsgroups by
people who claim to be expert coders?

Good development team has people who understand all this stuff -
physicists (or chemist, or biologists depending what the application
area is) as well
as coders. Ever heard of system engineers, applications engineers
(usually Ph.D. in the field), firmware engineers? Obviously everyone
has
its own role, and the designing complex software like one
we are talking about involves tens to hundreds of people.

There will not be any good people doing that work because you have
eliminated all danger. Nobody of that ilk will be interested in
that kind of work because it will be boring.

Does not mean there will be no mistakes. They will be corrected,
and each one needs to be corrected only once and for all million of
machines out there. At the other hand I observe tendency of
overestimating how good humans actually are.
If you look closely at any area of human activity, in average they
suck at it. And suck for good physical reasons, not only because of
faulty software (which is also widely present).

Right. People's efforts suck so much that you are posting based on
the work of ~100K people over the last 50 years. No computer can
innovate. What you want to do is destroy all opportunities to
innovate. This will stop trade. That will call a complete collapse
of the civilization you are familiar with. Perhaps this is a good
thing because it will eliminate all of these kinds of stupidos.

And you have to painstakingly try to fix this faulty software in
every single case individually, using very indirect and not always
successful methods (such as defensive driving school, fines, jail),
while with software you can just make a field update for all devices
in the world simultaneously in few minutes.

Man! Are you way out in la-la land. Computer: "Excuse me while
I reboot. Pay no attention to that pile of metal, blood, flesh,
and fire one mile ahead....system restarting....system restarting...
system restarting...".

What happens when the power grid shuts down?

Furthermore, there are vehicles
which have a need to break traffic laws such as police cars,
firetrucks, ambulances, and the rare times when a passenger
automobile has to break the usual laws due to an >emgergency.

Automatic vehicles only need to follow rules themselves, they would
not expect other vehicles to follow the rules.
This is a good practice for human drivers as well.
So cops can go ahead and break as many rules as they want.

Nope. With your scheme, no humans will be allowed to drive.


Finally, have you never heard of the acronym BSOD?

You don't even know what you don't know about how
computers work...or rather, don't work.

Every real time firmware has watchdog reset implemented on hardware
interrupt level. I know because I design such firmware.

I sure hope I don't ride transportation using your firmware designs.
You cannot issue a reset in the middle of a catastrophe if the
situation requires the hard/software to keep running. There are
other situations where the hard/software should stop running leaving
the controls and decision making to the human.

I don't know
why M$ people did not think of that...
Besides, if you look at system design of most DARPA vehicles, they
are running multiple computers in parallel

and what is the delay to resolve a conflict? How much time does
it take those parallel computers to decide that the conflict cannot
be resolved? When this happens, which of those parallel computers
becomes the primary CPU?

on separate interrupts and similar but not identical inputs, which
makes it less likely that a space particle
or Turing's theorem will make all of them hang simultaneously.

You should really think about deadly embraces and CATCH-22s. At
least with a simultaneous hang, the human will know to begin
making the decisions...if he has any time left.


And
even if they did, it goes into safe mode and simple low level
controller takes over and safely parks the vehicle on the road-side.

and if there is no side of the road? What if there is already a pile
of cars and semis there? What if the Grand Canyon is there?

And yes, one out of million, or out of 10 million will crash.

Oh, bullshit.

But
now we have 1 out of 77 crash (that is btw the life-time odds of dying
from transport related accident).

Where in the universe did you get those odds?

So one out of million would be a
tremendous improvement.


Assuming your numbers are correct, that one in a million crash will
not be a single car accident but a pile of cars, trucks, loads, and
fauna.


Bottom line - bugs will be there, but human hardware and software has
bugs too, only these are not fixable.

The coders are people. So you are stating that those who write the
bugs cannot fix them and will never learn from their
mistakes.



We are not choosing perfection
vs. imperfection, we are just choosing better of the two
imperfections.

Oh, nuts. You are choosing dumbing down all people and
a government who will dictate who will be allowed to
go where and when.

/BAH

.

Relevant Pages

  • Re: Driverless Car ?!
    ... Assuming that there would be fewer crashes with automated cars on ... Y2K bugs. ... Automatic vehicles only need to follow rules themselves, ... I know because I design such firmware. ...
    (sci.physics)
  • Re: The Cult Of Mac
    ... Beer, sports, cars, computers, power tools, game ... I don't see people buy different kids of skateboards ... There are Mac zealots just as ...
    (comp.sys.mac.advocacy)
  • Re: Google Self-Driiving Car
    ... and drives itself into the ditch and lets ... The strategy should be to give the computers driving the car even more ... It's cars. ... maintenance and certification is required on those components? ...
    (rec.autos.driving)
  • Re: Demand That Microsoft Sell No Code Before Its Time
    ... If developers cannot find the bugs themselves, ... over inefficient producers. ... the way "bugs" in cars were fixed was by ... But, the same is not true of software, as consumers continue ...
    (microsoft.public.windowsxp.general)
  • Re: SP2 - Why bother?
    ... Computers and cars are products, consumable items. ... hotfixes or Service packs. ...
    (microsoft.public.windowsxp.general)