Google Self-Driiving Car

News item this morning is that Google has had a car on the road for 2
years that drives itself. It has driven 142,000 miles without error.

The question is, would you want one? Consider how the speed would be
set. Do you want it to be out there, tied to the speed limit, when on
the same roads you have 18 wheelers catching you from behind at 75 -
80 mph? That's the situation on a lot of roads. Big trucks, and the
other traffic in general, are often waaaaaaaay above the speed limit.
The speed limit on I-95 approaching Washington, DC is 60 mph, but the
traffic flows in the high 70's most of the time. Its 55 on the
Washington, DC beltway, but 80 mph isn't rare.

Wanna go for a drive? Wanna do it in a "self-driving car" that is
going to get flattened by the 1st sleepy truck driver (they're all
sleepy, BTW.)

On Sun, 10 Oct 2010 08:53:24 -0400, Dave Head wrote:

> News item this morning is that Google has had a car on the road for 2
> years that drives itself. It has driven 142,000 miles without error.
>
> The question is, would you want one? Consider how the speed would be
> set. Do you want it to be out there, tied to the speed limit, when on
> the same roads you have 18 wheelers catching you from behind at 75 -
> 80 mph? That's the situation on a lot of roads. Big trucks, and the
> other traffic in general, are often waaaaaaaay above the speed limit.
> The speed limit on I-95 approaching Washington, DC is 60 mph, but the
> traffic flows in the high 70's most of the time. Its 55 on the
> Washington, DC beltway, but 80 mph isn't rare.
>
> Wanna go for a drive? Wanna do it in a "self-driving car" that is
> going to get flattened by the 1st sleepy truck driver (they're all
> sleepy, BTW.)

http://www.pcmag.com/article2/0,2817,2370518,00.asp

No mention of the cars being on a real interstate for any length of time
other than to cross the golden gate bridge.
The mileage is not from one car, but several.

On Sun, 10 Oct 2010 10:34:35 -0400, richard >
wrote:

>On Sun, 10 Oct 2010 08:53:24 -0400, Dave Head wrote:
>
>> News item this morning is that Google has had a car on the road for 2
>> years that drives itself. It has driven 142,000 miles without error.
>>
>> The question is, would you want one? Consider how the speed would be
>> set. Do you want it to be out there, tied to the speed limit, when on
>> the same roads you have 18 wheelers catching you from behind at 75 -
>> 80 mph? That's the situation on a lot of roads. Big trucks, and the
>> other traffic in general, are often waaaaaaaay above the speed limit.
>> The speed limit on I-95 approaching Washington, DC is 60 mph, but the
>> traffic flows in the high 70's most of the time. Its 55 on the
>> Washington, DC beltway, but 80 mph isn't rare.
>>
>> Wanna go for a drive? Wanna do it in a "self-driving car" that is
>> going to get flattened by the 1st sleepy truck driver (they're all
>> sleepy, BTW.)
>
>http://www.pcmag.com/article2/0,2817,2370518,00.asp
>
>No mention of the cars being on a real interstate for any length of time
>other than to cross the golden gate bridge.
>The mileage is not from one car, but several.

Interesting. The difference between reading a magazine article and
getting it from a TV news story. The magazine's going to have more
details.

The auto companies have projected that we would have self-driving cars
by 2020. Well, I'm all for it, but that's a long way to go.

Other considerations are whether the cars are going to be programmed
with CNN Money's evaluation of the top 25 most dangerous neighborhoods
and avoid those, or not do like the German guy a few years ago that
followed his GPS right into a river because the ferry boat was not at
the dock. Stuff like that. How smart are these cars going to be?
How smart are computers going to get? Maybe they'll be better than
us, and have a network that detects an earthquake miles away that
stops our cars from driving across a bridge as the earthquake strikes.
This could be very very good, or fairly bad.

On Sun, 10 Oct 2010 12:49:16 -0400, Dave Head wrote:

> On Sun, 10 Oct 2010 10:34:35 -0400, richard >
> wrote:
>
>>On Sun, 10 Oct 2010 08:53:24 -0400, Dave Head wrote:
>>
>>> News item this morning is that Google has had a car on the road for 2
>>> years that drives itself. It has driven 142,000 miles without error.
>>>
>>> The question is, would you want one? Consider how the speed would be
>>> set. Do you want it to be out there, tied to the speed limit, when on
>>> the same roads you have 18 wheelers catching you from behind at 75 -
>>> 80 mph? That's the situation on a lot of roads. Big trucks, and the
>>> other traffic in general, are often waaaaaaaay above the speed limit.
>>> The speed limit on I-95 approaching Washington, DC is 60 mph, but the
>>> traffic flows in the high 70's most of the time. Its 55 on the
>>> Washington, DC beltway, but 80 mph isn't rare.
>>>
>>> Wanna go for a drive? Wanna do it in a "self-driving car" that is
>>> going to get flattened by the 1st sleepy truck driver (they're all
>>> sleepy, BTW.)
>>
>>http://www.pcmag.com/article2/0,2817,2370518,00.asp
>>
>>No mention of the cars being on a real interstate for any length of time
>>other than to cross the golden gate bridge.
>>The mileage is not from one car, but several.
>
> Interesting. The difference between reading a magazine article and
> getting it from a TV news story. The magazine's going to have more
> details.
>
> The auto companies have projected that we would have self-driving cars
> by 2020. Well, I'm all for it, but that's a long way to go.
>
> Other considerations are whether the cars are going to be programmed
> with CNN Money's evaluation of the top 25 most dangerous neighborhoods
> and avoid those, or not do like the German guy a few years ago that
> followed his GPS right into a river because the ferry boat was not at
> the dock. Stuff like that. How smart are these cars going to be?
> How smart are computers going to get? Maybe they'll be better than
> us, and have a network that detects an earthquake miles away that
> stops our cars from driving across a bridge as the earthquake strikes.
> This could be very very good, or fairly bad.

Hell, the tv moguls all said in the early fifties that in 1964 driving a
car a long distance would mean hooking up to an automated system and
letting the system do the driving for you.

While there are self driving vehicles at work, they don't go much of
anywhere yet but on that property, like at airports.

The technology is here to do it on a widescale, but it will be decades and
a couple of generations before that idea catches on and becomes the thing.
It ain't gonna happen because the automotive industry won't let it happen.

Dave Head > wrote in
news
> On Sun, 10 Oct 2010 10:34:35 -0400, richard >
> wrote:
>
>>On Sun, 10 Oct 2010 08:53:24 -0400, Dave Head wrote:
>>
>>> News item this morning is that Google has had a car on the road for 2
>>> years that drives itself. It has driven 142,000 miles without error.

the article said a FLEET of cars.

"without error"? how did they measure that?

>>>
>>> The question is, would you want one? Consider how the speed would be
>>> set. Do you want it to be out there, tied to the speed limit, when on
>>> the same roads you have 18 wheelers catching you from behind at 75 -
>>> 80 mph? That's the situation on a lot of roads. Big trucks, and the
>>> other traffic in general, are often waaaaaaaay above the speed limit.
>>> The speed limit on I-95 approaching Washington, DC is 60 mph, but the
>>> traffic flows in the high 70's most of the time. Its 55 on the
>>> Washington, DC beltway, but 80 mph isn't rare.
>>>
>>> Wanna go for a drive? Wanna do it in a "self-driving car" that is
>>> going to get flattened by the 1st sleepy truck driver (they're all
>>> sleepy, BTW.)
>>
>>http://www.pcmag.com/article2/0,2817,2370518,00.asp
>>
>>No mention of the cars being on a real interstate for any length of time
>>other than to cross the golden gate bridge.
>>The mileage is not from one car, but several.
>
> Interesting. The difference between reading a magazine article and
> getting it from a TV news story. The magazine's going to have more
> details.
>
> The auto companies have projected that we would have self-driving cars
> by 2020. Well, I'm all for it, but that's a long way to go.

I'm "all for it" for those who have had their licenses revoked,or poor
driving records.
Otherwise,not so much. I LIKE to drive.
Plus,there's what could happen with a SW glitch or circuit failure.
Muffy is texting on her cellphone or watching a DVD,and absolutely clueless
as to what's going on on the road. No way she could "retake control" in any
useful time frame.

also,how does their SW do for AVOIDANCE of other drivers errors?
Will it keep me out of a collision? Or will it send me and my car into a
ditch or body of water?
>
> Other considerations are whether the cars are going to be programmed
> with CNN Money's evaluation of the top 25 most dangerous neighborhoods
> and avoid those, or not do like the German guy a few years ago that
> followed his GPS right into a river because the ferry boat was not at
> the dock. Stuff like that. How smart are these cars going to be?
> How smart are computers going to get? Maybe they'll be better than
> us, and have a network that detects an earthquake miles away that
> stops our cars from driving across a bridge as the earthquake strikes.
> This could be very very good, or fairly bad.
>

I suspect very bad.
Plus,what will it do to the PRICE of a car? Those who could afford it could
afford to hire chauffeurs.

--
Jim Yanik
jyanik
at
localnet
dot com

On Sun, 10 Oct 2010 16:30:03 -0500, Jim Yanik >
wrote:

>I'm "all for it" for those who have had their licenses revoked,or poor
>driving records.

Sounds good.

>Otherwise,not so much. I LIKE to drive.

Ditto.

>Plus,there's what could happen with a SW glitch or circuit failure.

One word: hackers...

--
"I... Can't drive... FIFTY-FIVE!!"
--Sammy Hagar

On Mon, 11 Oct 2010 21:52:57 +0800, Bernd Felsche
> wrote:


>Solve the problem: Idiots on the roads.

There is no solution for this. Everyone has to drive in this country
or the economy would collapse.

>>Accident avoidance will have to be part of the programming, or it
>>won't sell.
>
>It'll be mandated because "it saves lives".

Accident avoidance is already mandated, since I wouldn't buy a
self-driving car that didn't have it. I don't think you would either.
Its called "defensive driving" when humans do it, and without it, you
probably won't live very long.

>Politics turns suppositions into truths.

>Mandatedjust like you can't drive a classic car in major European
>cities any more because those cars "damage the environment" by
>emitting nothing more than harmless plant food.

Well... its a bit more than that, but, the rule is ridiculous, of
course.

>An autonomous vehicle system that assesses all the risks will figure
>out that there's a finite probability of an accident. It may not
>necessarily refuse to depart on a journey, but the circumstances of
>a journey could suddenly induce a "catatonic fit" where the
>perceived situation becomes too "dangerous" and the system can't
>figure out how to deal with it gracefully.

I'm thinking more along the lines of the vehicle detecting the sleepy
trucker about to hit me from behind while I'm stopped for a road
blockage on an interstate, and drives itself into the ditch and lets
the truck hit someone else. Things like that.

>If you don't believe that that is a real situation; consider the
>paradox of the intersection; where the traffic on the priority road
>expects to flow freely. Recall that when travelling on such a road,
>that there is inevitably a vehicle waiting to enter/cross from a
>side road. If that car's operator doesn't follow the rules, the
>probability of a crash reaches 1 when the driver on the priority
>road is about a quarter of a second away.

A computer-controlled car would, I expect, be watching for the
slightest indication of movement from such a car, and be prepared to
take evasive action. The thing that comes to mind that would likely
be most effective would be to sound the horn.

>And there is no way that the robot can tell for sure IF the driver
>of the car on the side road will be playing by the rules; or has
>perceived the vehicle on the priority road.

I can do it. I figure a computer with TV cameras and maybe radar can
do it too.

>And then there is the situation of accident impact amelioration if
>the autonomous vehicle is programmed to take risks (Do you feel
>lucky?).

Driving as a whole is a risk. Its just how much risk you're willing
to take. Some smart guy once said that Injury is the price of
movement, which is true. The goal is to minimize the injury.

>When the unanticipated happens, what actions will minimise not only
>the chance of injury to the vehicle's occupants, but also those in
>the traffic environment.

Yep.

>People who *drive* intrinsically understand these things. Without
>knowing hard rules. A judgement, based on very little active
>information is made in a split second to decide where and how to
>crash; if one must crash.

The strategy should be to give the computers driving the car even more
information than a person has. If things do go south, the computer
should be programmed to miss the trees.

>>Price shouldn't be a problem, since this is computers. Computers get
>>cheaper all the time. Software is the big bug-a-boo with computers,
>
>No it isn't computers. It's cars. Computer hardware, actuators and
>sensors for automotive environments is very much more expensive than
>for disposable consumer goods.

Cars are going to be expensive whether they have computers or not. The
computers, I'm saying, won't add much to the price of the cars. The
software might, but if its built for millions of cars, then the price
shouldn't be all that much. Figure the software as a 10 year project
for 1000 engineers, each making $100K / yr. That's a billion dollars.
GM was, at one time, making 10 million cars a year. Use the software
in all those cars, and charge $100 / car for the software, and...
you're done. Everything after that is profit.

>Are the avionics in aircraft becoming cheaper? What level of
>maintenance and certification is required on those components?

What's the economy of scale for aircraft? Will cars get into
"certification?" They have software in Toyotas right now controlling
throttles that, in all likelihood, goes wacko on rare occasionsw and
wide-opens the throttle. Nobody's calling for certification, and
certification wouldn't find such a rare problem anyway.

>Tell me which bit of non-trivial firmware is identical in millions
>of cars.

I didn't say that existed, just that it ought to, for such a project.
In fact, it should be open source, so anyone can see the code, and
search for bugs in it as a hobby if they want to.

>Any software errors in vehicle guidance systems would make the cars
>unsafe to drive.

Yeah, like the Toyota throttle, perhaps?

>All vehicles with that software would have to be
>recalled, perhaps towed away, in order to fix the problem.

It might be a matter of someone showing up with a box and downloading
patched software, too, right along the road. Depends on the severity
of the error.

>If there is only one type of vehicle control software for all makes,
>then that creates severe sociological risks. Consider the problems
>of "everybody" having the same operating system on their computer.

>You can't solve the sociological problem of idiots on the road with
>technology.

You can if you keep 'em out of the control loop.

Dave Head > wrote in
:

> On Mon, 11 Oct 2010 21:52:57 +0800, Bernd Felsche
> wrote:
>
>
>>Solve the problem: Idiots on the roads.
>
> There is no solution for this. Everyone has to drive in this country
> or the economy would collapse.

a defeatist attitude. actually contributing to the problem.
To assume that "nothing can be done" for fear of "X" is accepting bad
driving.
>
>>>Accident avoidance will have to be part of the programming, or it
>>>won't sell.
>>
>>It'll be mandated because "it saves lives".
>
> Accident avoidance is already mandated, since I wouldn't buy a
> self-driving car that didn't have it. I don't think you would either.
> Its called "defensive driving" when humans do it, and without it, you
> probably won't live very long.
>
>>Politics turns suppositions into truths.
>
>>Mandatedjust like you can't drive a classic car in major European
>>cities any more because those cars "damage the environment" by
>>emitting nothing more than harmless plant food.
>
> Well... its a bit more than that, but, the rule is ridiculous, of
> course.
>
>>An autonomous vehicle system that assesses all the risks will figure
>>out that there's a finite probability of an accident. It may not
>>necessarily refuse to depart on a journey, but the circumstances of
>>a journey could suddenly induce a "catatonic fit" where the
>>perceived situation becomes too "dangerous" and the system can't
>>figure out how to deal with it gracefully.
>
> I'm thinking more along the lines of the vehicle detecting the sleepy
> trucker about to hit me from behind while I'm stopped for a road
> blockage on an interstate, and drives itself into the ditch and lets
> the truck hit someone else. Things like that.

I can envision situations where the program decides that your car cannot
move at all as part of the "avoidance" module.
Maybe the weather is too bad,or some other reason. Maybe a screwy sensor is
telling the program there's an obstruction or hazard.
>
>>If you don't believe that that is a real situation; consider the
>>paradox of the intersection; where the traffic on the priority road
>>expects to flow freely. Recall that when travelling on such a road,
>>that there is inevitably a vehicle waiting to enter/cross from a
>>side road. If that car's operator doesn't follow the rules, the
>>probability of a crash reaches 1 when the driver on the priority
>>road is about a quarter of a second away.
>
> A computer-controlled car would, I expect, be watching for the
> slightest indication of movement from such a car, and be prepared to
> take evasive action. The thing that comes to mind that would likely
> be most effective would be to sound the horn.

I can see several of these robocars at an intersection,when one moves the
others halt,and then the first car halts because one of the others began a
move. how many interacting situations could an affordable computer handle
at one time?
>
>>And there is no way that the robot can tell for sure IF the driver
>>of the car on the side road will be playing by the rules; or has
>>perceived the vehicle on the priority road.
>
> I can do it. I figure a computer with TV cameras and maybe radar can
> do it too.

You are a dreamer,considering today's present technology level.
>
>>And then there is the situation of accident impact amelioration if
>>the autonomous vehicle is programmed to take risks (Do you feel
>>lucky?).
>
> Driving as a whole is a risk. Its just how much risk you're willing
> to take. Some smart guy once said that Injury is the price of
> movement, which is true. The goal is to minimize the injury.
>
>>When the unanticipated happens, what actions will minimise not only
>>the chance of injury to the vehicle's occupants, but also those in
>>the traffic environment.
>
> Yep.
>
>>People who *drive* intrinsically understand these things. Without
>>knowing hard rules. A judgement, based on very little active
>>information is made in a split second to decide where and how to
>>crash; if one must crash.
>
> The strategy should be to give the computers driving the car even more
> information than a person has. If things do go south, the computer
> should be programmed to miss the trees.

"should be".
and instead,it dives into a pond or lake and you sink,or a ditch,and you
roll over. Driving is a VERY complex problem,with MANY variables.
Any computer would have to be programmed with EVERY possible problem and
how to avoid it.At least people can reason and figure out what things are
important and what a likely course of actions would be.

you're talking artificial intelligence systems here.Able to recognize any
sort of threat or problem and figure the right course of action in
milliseconds.
VISAR or ZORAC could do it,but they are sci-fi,Ganymean technology.
(James P. Hogan's Giant's novels.)
>
>>>Price shouldn't be a problem, since this is computers. Computers get
>>>cheaper all the time. Software is the big bug-a-boo with computers,
>>
>>No it isn't computers. It's cars. Computer hardware, actuators and
>>sensors for automotive environments is very much more expensive than
>>for disposable consumer goods.
>
> Cars are going to be expensive whether they have computers or not. The
> computers, I'm saying, won't add much to the price of the cars. The
> software might, but if its built for millions of cars, then the price
> shouldn't be all that much. Figure the software as a 10 year project
> for 1000 engineers, each making $100K / yr. That's a billion dollars.
> GM was, at one time, making 10 million cars a year. Use the software
> in all those cars, and charge $100 / car for the software, and...
> you're done. Everything after that is profit.
>
>>Are the avionics in aircraft becoming cheaper? What level of
>>maintenance and certification is required on those components?

Pilots are still the ones flying the plane.
and air traffic is nowhere near as dense as auto traffic.
I note commercial passenger aircraft still don't allow TCAS to steer the
plane when a near miss or collision is imminent.It just gives a
warning,some info,and the pilot makes decisions and acts.
>
> What's the economy of scale for aircraft? Will cars get into
> "certification?" They have software in Toyotas right now controlling
> throttles that, in all likelihood, goes wacko on rare occasionsw and
> wide-opens the throttle. Nobody's calling for certification, and
> certification wouldn't find such a rare problem anyway.

Insurance companies are sure to call for certification.
>
>>Tell me which bit of non-trivial firmware is identical in millions
>>of cars.
>
> I didn't say that existed, just that it ought to, for such a project.
> In fact, it should be open source, so anyone can see the code, and
> search for bugs in it as a hobby if they want to.
>
>>Any software errors in vehicle guidance systems would make the cars
>>unsafe to drive.
>
> Yeah, like the Toyota throttle, perhaps?
>
>>All vehicles with that software would have to be
>>recalled, perhaps towed away, in order to fix the problem.
>
> It might be a matter of someone showing up with a box and downloading
> patched software, too, right along the road. Depends on the severity
> of the error.

again the Utopian dreaming.
>
>>If there is only one type of vehicle control software for all makes,
>>then that creates severe sociological risks. Consider the problems
>>of "everybody" having the same operating system on their computer.
>
>>You can't solve the sociological problem of idiots on the road with
>>technology.
>
> You can if you keep 'em out of the control loop.
>



--
Jim Yanik
jyanik
at
localnet
dot com

On Mon, 11 Oct 2010 14:01:13 -0500, Jim Yanik >
wrote:

>a defeatist attitude. actually contributing to the problem.
>To assume that "nothing can be done" for fear of "X" is accepting bad
>driving.

>I can envision situations where the program decides that your car cannot
>move at all as part of the "avoidance" module.
>Maybe the weather is too bad,or some other reason. Maybe a screwy sensor is
>telling the program there's an obstruction or hazard.
>
>I can see several of these robocars at an intersection,when one moves the
>others halt,and then the first car halts because one of the others began a
>move. how many interacting situations could an affordable computer handle
>at one time?

>You are a dreamer,considering today's present technology level.

>"should be".
>and instead,it dives into a pond or lake and you sink,or a ditch,and you
>roll over. Driving is a VERY complex problem,with MANY variables.
>Any computer would have to be programmed with EVERY possible problem and
>how to avoid it.At least people can reason and figure out what things are
>important and what a likely course of actions would be.
>
>you're talking artificial intelligence systems here.Able to recognize any
>sort of threat or problem and figure the right course of action in
>milliseconds.
>VISAR or ZORAC could do it,but they are sci-fi,Ganymean technology.
>(James P. Hogan's Giant's novels.)

>Pilots are still the ones flying the plane.
>and air traffic is nowhere near as dense as auto traffic.
>I note commercial passenger aircraft still don't allow TCAS to steer the
>plane when a near miss or collision is imminent.It just gives a
>warning,some info,and the pilot makes decisions and acts.

>Insurance companies are sure to call for certification.
>
>again the Utopian dreaming.

I think you shouldn't buy one. Just don't get in my way if they come
out in 2020 as predicted. I'm really, really tired of all the f'n
idiots on the road, and about a 50-50 of me getting blamed for it if
anything goes wrong. That's the way they operate around here. If
there's an accident, whoever survives gets thrown in jail. If nobody
survives, they look around for the nearest bystanders. Would be real
glad not to be in that line of fire any more.

BTW, if you engineer a car the way you expect those to be engineered,
I wouldn't get in one you built either.

On Mon, 11 Oct 2010 13:38:53 -0500, Jim Yanik >
wrote:

>What's the COST? what if the sensor craps out? or is just blocked by ice or
>dirt?

>SO WHAT? it's not a matter of computing power.


>You see things in a very Utopian manner.
>Perhaps insurance companies,not having any base DATA to use to calculate
>their risk factors,will charge a huge premium for a car with such a
>system,at least until the system proves itself in real-life conditions.
>Insurance companies are pretty smart,risk averse,and cost-conscious.
>
>BTW,HOW does a drunk "take over" if the HAL9000 goes awry?

>No,the bad drivers would either change or suffer.
>We don't have to accept bad driving. If they can't get to work,then the
>companies hire people who can. They do that now.

>They can car-pool,taxi,take public transpo,ride a bike,or walk.

You're not making much sense. You say its not a matter of computing
power and state elsewhere that there's a lot of graphics involved.
Graphics are nothing but computing power. You haven't seen the size of
the program, but that much AI is not going to be simple, and will have
to be really parallel to be really quick. Absolutely it'll be a
computing monster to make it happen. The car companies are talking
2020. I dunno. The make it, I'll buy it. I'll be 73 by that time
anyway, if I'm really lucky, and just 10 years sicker of driving.