Google Spin-Off's Newest Self-Driving Minivans Start Road Tests This Month

"We're at an inflection point where we can begin to realize the potential of this technology," the CEO of Waymo said today. An anonymous reader quotes The Verge: Waymo, the self-driving car startup spun-off from Google late last year, will be deploying its fleet of self-driving Chrysler Pacifica minivans onto public roads for the first time later this month, the company announced at the North American International Auto Show. The minivans will be hitting the roads in Mountain View, California and Phoenix, Arizona, where the company's self-driving Lexus SUVs have already driven thousands of miles over the past few years...

But here's the thing about these minivans. Waymo says that for the first time, it's producing all the technology that enables its cars to completely drive themselves in-house... This allows the company to exert more control over its self-driving hardware, as well as bring the cost down to ridiculously cheap levels. In a speech in Detroit, Waymo CEO Jeff Krafcik said that by building its own LIDAR sensors, for example, the company was shaving 90 percent off its costs. That means sensors that Google purchased for $75,000 back in 2009 now only cost $7,500 for Waymo to build itself.
Waymo's CEO says that using high-resolution LIDAR sensors "helps us more accurately predict where someone will walk next."

Talking heads talking

[Kjella]

That means sensors that Google purchased for $75,000 back in 2009 now only cost $7,500 for Waymo to build itself.

I'm guessing cost was waaaaaay down on Google's list of priorities when they picked experimental sensors for their highly experimental cars. I doubt anyone thought it had to be these exact sensors that go on production models. Also this has very little to do with in-house, since 2009 there are tons [dronezon.com] of new, much cheaper LIDAR sensors. Minus the PR blurb it's basically "Waymo follows industry, pretends it's a miracle".

Re:

[DreadCthulhu]

This could be the plug-in hybrid version of the Pacifica, which has a 33 mile electric range, and is rated at 32 mpg combined when the battery runs out. That is vastly better than any competing minivan or SUV that can carry a similar number of people.

Re:Minivans are terrible for the environment

[CrankyFool]

In fact, if you look at the article, and then watch the video, and then wait about a minute, you'll see the minivans they show are all plug-in electrics. (This isn't a "why the hell won't people read the article?" complaint -- it takes way too much time to figure this out. The only reason I bothered is because my spouse is deeply interested in having the Pacifica electric minivan as their next car, so I'm an interested party)

Competition is heating up

[djinn6]

Nice to see even more competition in this space. This has gone from the realm of science fiction to (soon to be) reality.

With LIDAR, these cars should see better than any human possibly can and shouldn't run into the problem of differentiating clouds and trucks, or be blinded by the sun, rain or fog. So all that's left is the software, and Google's been working on that for ages. Of all the self-driving cars around, I think this is the closest to a truly self-driving car, one that doesn't need a driver to babysit it constantly.

And before all the cynics start moaning about bad weather performance, let me just remind everyone how well humans drive in the snow [youtube.com].

Re:

[fluffernutter]

I've heard lidar is blinded by fog. Also it needs to stop for every shopping bag blowing in the wind, and they don't seem to be able to figure out where to mount them so that they can detect everything from roof level to a few inches off the ground.

Re:Competition is heating up

[ShanghaiBill]

I've heard lidar is blinded by fog.

Correct. So when the car encounters fog, it should slow down. Just like a human would. But Google cars also have radar, which can penetrate fog, so they can still do better than humans.

Also it needs to stop for every shopping bag blowing in the wind

The "plastic bag problem" was solved a couple years ago, using ANNs. The vision system can differentiate between a rock and a plastic bag as well as a human that is alert and paying attention, and do way better than humans yakking on their cellphones or yelling at the kids in the backseat.

Re:

[Rothron the Wise]

Even with Radar you probably want to slow down, because that's what the other drivers expect.

Re:

[fluffernutter]

Ok well I have yet to see or hear about it working together correctly. The last article I saw, they needed to install a special lidar sensor so the car wouldn't run over squirrels. It makes me wonder how many sensors we will need to cover a car with just to see everything that a car needs to see in order to drive without coming into contact with, or running over anything. Humans aren't perfect at avoiding squirrels all the time either, I understand that, but at least they try and an AI car is supposed to

Re:

[sinij]

And before all the cynics start moaning about bad weather performance, let me just remind everyone how well humans drive in the snow [youtube.com].

For this specific example (total loss of traction on 4 wheels) what makes you think self-driving car would do any better?

Re:

[ThatsMyNick]

Probably nothing makes him think self driving car would do better in that scenario. Not sure where you got that from. His point is self driving cars would be no worse than humans in bad weather conditions. It may be better in some cases (it may refuse to start driving in such bad weather, avoiding routes with bad weather for long distance trips), worse in some cases too, overall I dont think it would be a lot better.

Re:

[fluffernutter]

it may refuse to start driving in such bad weather,

Is it going to go do my job for me? Because my boss won't give a shit whether my car refuses to drive me there.

Re:

[ThatsMyNick]

No, if it is unsafe for the car to drive itself, it is unsafe for you to drive too. I am sure your boss will prefer you stay home when roads are filled with black ice and snow. You should be looking for a new boss, if not.

Re:

[fluffernutter]

So basically if the car decided to drive walking speed or stop and back up all the humans behind it trying to get somewhere, you're willing to say "that's ok" because a machine says it is so? What a depressing world we are headed for.

Re:

[fluffernutter]

No, if it is unsafe for the car to drive itself, it is unsafe for you to drive too.

That's what he is basically saying. How else should I take it? He's saying I should not drive if an AI decides not to drive.

Re:

[ThatsMyNick]

No, that is not what I said. The car will decide when you give it the destination (so presumably your drive way or parking lot or whatever). In rare cases if it notices something only when driving, it will do whatever a human will do when they notice black ice when driving (drive slowly, safely, or if that is impossible, yeah just block all cars in the road, which is safer for everyone right? If you think the machine is malfunctioning, you call a tow, just like you would when your car is malfunctioning). Ye

Re:

[fluffernutter]

My concern is that these machines will just start driving walking speed everywhere since the company making the machine will be afraid of being sued, won't really have the right equipment to drive on ice, so where a human would get to work in 30 minutes safely it will now take 2 hours because there are no assurances these machines will decide to drive as well as a human in all conditions. Traffic will be even more clogged than it already gets on an icy, snowy morning. That's depressing to think about.

It

Re:

[ThatsMyNick]

They wont be adopted if the machine is being unreasonable. Not that depressing outcome. I dont see why it would be unreasonable though. Insurance companies expect pile ups everytime there is bad weather already.

Re:

[fluffernutter]

I'm not so sure about that. People on Slashdot seem to want to forge ahead with these things no matter how they are working. As long as there are enough people to buy these things that will happily watch a movie while humans pile up behind them to make them profitable, companies seem to feel entitled to sell them. The only thing that stops companies is the law, and there are no minimum speed limits, so... *sigh*

Re:

[fluffernutter]

I mean, a company will get sued into oblivion if they kill someone due to a fault in programming, so why wouldn't they drive slower than they need to?

Re:

[ThatsMyNick]

All it takes in one self driving car to cause a jam, likewise all it takes is one over confident idiot to cause a jam. I dont think it would be that different to be honest.

And no, there wont be a minimum speed limit. If a bunch of people (or even a single person) decides 5 miles per hour is the speed for the conditions, 5 mph it will be.

You have too much faith in the present. I am as pessimistic about the present and the future.

Re:

[fluffernutter]

Except most self driving cars will drive the same way, so if you have 1000 self driving cars on the road and one automaker drives slow, you have 300 self driving cars driving slow. I don't know why people here seem to think humans are so bad at driving, I see thousands of humans on the road every day and only the odd one doing anything wrong.

Re:

[ThatsMyNick]

I have seen humans drive in bad weather (I had to drive across Colorado in one), humans are horrible at unexpected and unprepared situations.

Re:

[fluffernutter]

Then I guess I'm just imagining seeing people driving down the road and not crashing.

Re:

[fluffernutter]

Any way, that's not the point. The point is that I feel AI automakers are not prepared to accept responsibility if they make the problems worse.

Re:

[fluffernutter]

"As well as" of course means "as safely as" and "as quickly as".

Re:

[ShanghaiBill]

For this specific example (total loss of traction on 4 wheels) what makes you think self-driving car would do any better?

Because the SDC is more likely to be driving slower when road conditions are bad. Also the SDC can use vehicle-to-vehicle communication to learn about black ice 200 meters ahead.

Re:

[RivenAleem]

I wonder at what point the cars will have to blank out the windscreens. When fog is no impediment to these things, and they are driving at normal (when normal may ultimately be a lot faster than we currently do) in 0 visibility, will passengers be able to handle it?

Also, are windows a weak-point in cars, and will removing them altogether to make a more robust vehicle be the next step?

Whamo!

[Mr. Droopy Drawers]

When I heard about this on the radio last night, I kept hearing "Whamo!" [wikipedia.org] instead of "Waymo". Kind of a bad name for self-driving car tech.