Self Driving Cars?

[JoeWras]

I think a Waymo when becomes available for sale to public will be way mo expensive than what I would pay.

... Queue up and play the rim shot sound effect ...

 

[ERD50]

Here, a reporter rode in a Waymo car, which still had a backup human driver.

The passenger riding in the car was shown what the car sensors saw, such as other cars on the road as well as pedestrians crossing the road.

I hadn't noticed those rotating (oscillating?) scanners (radar? lidar is on top?) at the front and corners before. see at 42 second mark.

The seem so exposed. Is that necessary? That's not gonna work in snow/ice!

-ERD50

 

[NW-Bound]

These are all part of the sensor suite deployed from the beginning on these Pacificas, and announced one year ago in Jan 2018 by Waymo CEO John Krafcik at an auto convention. I described them before in a long-ago post on this thread.

The small auxiliary lidars at the 4 corners of the cars, and at the front and rear are meant for close object detection. That's what you need to avoid running over or backing over little kids on tricycles, dogs, cats, etc..., or parking curbs, and what have you. There are also sonars and cameras at these corners too.

The main lidar and optical cameras are on the dome on the roof top. These look out several hundred feet for high-speed driving and cannot look down, hence need to be augmented by close-range sensors.

That's what you need for a truly safe SDC, if you do not want to run into freeway barriers, parked vehicles, etc..., or over pedestrians, bicyclists, and even your own kid on your driveway.

It takes a lot more than people realize. And these sensors will require maintenance. And that's why Waymo will not be selling these cars soon, but will operate a taxi fleet first.

 

[garyt]

I just hope they're up and running in 20-25 years when I'll likely be giving up driving.

 

[NW-Bound]

Seems like they have it running now or soon, but it takes "way mo" moolah for such a car, as ls99 noted.

Money to buy, and also to maintain such cars.

 

[NW-Bound]

Correction: The announcement of this in-house sensor suite by Krafcik, CEO of Waymo, was made back in early 2017, not 2018.

So, it has been 2 years, and they already had the sensors working at that point, with data to show in a public presentation way back then.

 

[JoeWras]

I just hope they're up and running in 20-25 years when I'll likely be giving up driving.

Move to dry, southern deserts and you'll be ahead of the curve.

 

[Midpack]

Move to dry, southern deserts and you'll be ahead of the curve.

+1. I haven’t seen any team demonstrate snow-ice-heavy rain capabilities yet, but obviously the teams realize the challenge.

 

[Walt34]

+1. I haven’t seen any team demonstrate snow-ice-heavy rain capabilities yet, but obviously the teams realize the challenge.

That'll be interesting to see them pull off in terrain like we have here. Lots of narrow roads with sharp drop-offs at the edges of the pavement. In heavy snow you'd better be a local or stay where you are, or you'll get stuck for sure, not knowing where the pavement edge is when the roadway takes a bend and the scenery does not, if that makes sense. The SDC would need a system to read through varying densities of snow (freshly fallen to hard packed by traffic) to "see" where the pavement is.

I'm sure it can be done but the price tag is probably going to be high.

 

[NW-Bound]

An SDC car that's smart enough would decline to drive in really bad inclement weather. It would say "Nope. You go kill yourself".

But that would mean you cannot take away the steering wheel and pedals.

 

[JoeWras]

I've been a skeptic on the timeline, not necessarily the ultimate technology.

But I'm serious about the dry, warm weather thing. I could see "zones" be approved for SDCs in places like Phoenix that are laid out in nice, modern, grids where the weather is generally favorable. And in the times when rain or or dust or other bad weather comes, the SDCs may be "grounded."

Initially I thought SDCs may be approved on specific highways. I'm coming around to specific areas where conditions are favorable, and the municipalities welcoming.

 

[ERD50]

These are all part of the sensor suite deployed from the beginning on these Pacificas, and announced one year ago in Jan 2018 by Waymo CEO John Krafcik at an auto convention. I described them before in a long-ago post on this thread.

The small auxiliary lidars at the 4 corners of the cars, and at the front and rear are meant for close object detection. That's what you need to avoid running over or backing over little kids on tricycles, dogs, cats, etc..., or parking curbs, and what have you. There are also sonars and cameras at these corners too.

The main lidar and optical cameras are on the dome on the roof top. These look out several hundred feet for high-speed driving and cannot look down, hence need to be augmented by close-range sensors.

That's what you need for a truly safe SDC, if you do not want to run into freeway barriers, parked vehicles, etc..., or over pedestrians, bicyclists, and even your own kid on your driveway.

It takes a lot more than people realize. And these sensors will require maintenance. And that's why Waymo will not be selling these cars soon, but will operate a taxi fleet first.

Thanks - I had not realized they had lidars at the corners, I had thought only one on the roof dome. So Tesla doesn't even have a single lidar, and Waymo has what 6-7? And still has a lot of development work ahead of them.

If multiple lidar is a requirement (it seems to be), that's really going to be a cost/maintenance issue. Maybe this is so far out that cost will be down a lot by then?

I'd be more than happy with an interim far more limited 'assistance' mode. One where I was in control, and the system acted as a 2nd, 3rd, 4th set of eyes, and made sure that I was paying attention, and alerted to things I might miss. That's happening to an extent today, and can go further, w/o needing to be 100% like a full SDC. Better to have "good/better" today, than to wait many years for "perfect". And we can still be working on "perfect" in parallel.

edit/add - and also a "summon" and "park it" mode - it would be nice to be able to summon the car in a parking lot, or tell it to go park. Low speed driving in those conditions should be relatively easy/safe.

-ERD50

 

[JoeWras]

T
I'd be more than happy with an interim far more limited 'assistance' mode. One where I was in control, and the system acted as a 2nd, 3rd, 4th set of eyes, and made sure that I was paying attention, and alerted to things I might miss. That's happening to an extent today, and can go further, w/o needing to be 100% like a full SDC. Better to have "good/better" today, than to wait many years for "perfect". And we can still be working on "perfect" in parallel.

-ERD50

Well, it seems we are kind of going that way! Each model year seems to bring out some new assist that is a bit more sophisticated than the last.

 

[NW-Bound]

Thanks - I had not realized they had lidars at the corners, I had thought only one on the roof dome. So Tesla doesn't even have a single lidar, and Waymo has what 6-7? And still has a lot of development work ahead of them...

Two years ago, I reported on post #352 of this thread about what sensors Waymo cars carry.

I repeatedly made the point that Tesla tried to do with as few sensors as possible.

And of course, Tesla cars ran into semi-trailers, highway barriers, fire trucks, etc... And people still say "oh, but Musk is so smart, and Tesla cars are so advanced..." To this day, Tesla cars still cannot see the traffic lights.

Compared to Waymo cars, Tesla self-driving capability is no more than a toy, an advanced cruise control.

Here's an except of my original post almost 2 years ago.

...
The 2nd video is most interesting, as Waymo CEO introduced their own in-house sensor suite. It is installed on a fleet of Pacifica that is being tested. It is what I encountered several times when walking around my neighborhood. I thought I saw a lot of stuff poking out from the car, but could not catch it.

Anyway, instead of just one roof-mounted medium-range lidar, they now have the following suite:

* Roof-mounted lidars: one long and one medium-range, 360-deg view
* Front down looking, two side-looking, rear looking (4 total small lidars)
* Roof-mounted radar, 360-deg view
* Side-looking radars, mounted at 4 corners.
* Vision camera, top mounted, 360-deg view
* Vision camera, front-looking, high-resolution

Holy moly! That's a lot of sensors. They have all 3 types of sensors: lidars, radars, vision cameras covering 360 degrees. The roof-mounted lidars are for driving, and they cannot look down to objects close to the car, hence the 4 smaller lidars. The car is not going to run over, or back over something that is close to it.

 

[Midpack]

I've been a skeptic on the timeline, not necessarily the ultimate technology.

But I'm serious about the dry, warm weather thing. I could see "zones" be approved for SDCs in places like Phoenix that are laid out in nice, modern, grids where the weather is generally favorable. And in the times when rain or or dust or other bad weather comes, the SDCs may be "grounded."

Initially I thought SDCs may be approved on specific highways. I'm coming around to specific areas where conditions are favorable, and the municipalities welcoming.

Yep. City centers where speeds are low and road markings/signs are plentiful and well maintained, interstate/highways with low traffic density, and good weather conditions. Snow-ice-heavy rain, poorly marked roads, and higher traffic density roadways (until all cars are SDC) may come (much) later? SDC will have to be rolled out in stages, it won’t be anywhere and everywhere in all conditions from the outset? And off road maybe never? Off road and/or classic manual driving may be restricted to ‘driving parks’ - but that’s beyond our lifetimes?

 

[ERD50]

... And people still say "oh, but Musk is so smart, and Tesla cars are so advanced..." To this day, Tesla cars still cannot see the traffic lights.

...

I had no idea that it didn't already see traffic lights! Now I see that is "coming real soon":

https://electrek.co/2018/12/09/tesla-autopilot-soon-traffic-lights-self-driving-elon-musk/

I guess even I'm a victim of the hype. I really thought detecting traffic lights and stop/go as allowed was something that had been out there a long time.

You can hardly even approach the word "Autopilot" with a 10,000 meter pole if you can't detect traffic lights! Heck, a 3 year old will say "go" when they see the light turn green!

-ERD50

 

[NW-Bound]

I had no idea that it didn't already see traffic lights! Now I see that is "coming real soon":

https://electrek.co/2018/12/09/tesla-autopilot-soon-traffic-lights-self-driving-elon-musk/

I guess even I'm a victim of the hype. I really thought detecting traffic lights and stop/go as allowed was something that had been out there a long time.

You can hardly even approach the word "Autopilot" with a 10,000 meter pole if you can't detect traffic lights! Heck, a 3 year old will say "go" when they see the light turn green!

-ERD50

That's why one must always view "demo video" with a healthy dose of skepticism.

You never know if there was a human outside the view of the camera doing some "coaching" to help the computer along, in order to make an impressive and convincing video for public consumption.

 

[ls99]

Maybe 40 years ago was at a manufacturer of some elctronics stuff. The demo item had lots winking blinking LED's. I asked them to open the box to see inside. They refused. I called it running on vaporware. They got offended.

 

[NW-Bound]

I forgot to say that a fully autonomous autopilot that can navigate surface streets without human help takes a whole lot more than just recognizing and reading traffic lights. Freeway driving is a lot easier, and even then Tesla cars have run into stopped vehicles and concrete barriers.

I think Tesla's autopilot will not be there anytime soon compared to Waymo cars on surface streets. And even if it has some of the capabilities, the reliability will be so poor that Tesla will never step forward to indemnify the users for any accident that happens with its use.

In other words, the users will always be responsible, because Tesla's autopilot will be permanently in "beta mode".

 

[Nick12]

I forgot to say that a fully autonomous autopilot that can navigate surface streets without human help takes a whole lot more than just recognizing and reading traffic lights. Freeway driving is a lot easier, and even then Tesla cars have run into stopped vehicles and concrete barriers.

I think Tesla's autopilot will not be there anytime soon compared to Waymo cars on surface streets. And even if it has some of the capabilities, the reliability will be so poor that Tesla will never step forward to indemnify the users for any accident that happens with its use.

In other words, the users will always be responsible, because Tesla's autopilot will be permanently in "beta mode".

Fox news recently reporting a driver in a Tesla car fast asleep behind the wheel while the car was on autonomous autopilot. This was in Las Vegas.

 

[explanade]

Recent video posted by Cruise Automation on their SDC on the streets of San Francisco.

Those are challenging situations even for human drivers.

 

[NW-Bound]

Cruise Automation was acquired by GM in 2016. Below is a photo of a Cruise Automation test car.

On the Web, several versions of the sensor suite could be seen in photos taken at different time. The one shown below is dated Nov 2018, hence quite recent.

Note the dual roof-mounted lidars, plus the two side lidars. There appears to be another lidar on the front grille.

 

[explanade]

Yeah for sure it's unwieldy. I've seen Apple testing these Lexus SUVs and it looks like they have a giant toboggan on the roof.

The Waymo minivans in Silicon Valley also have a lot of things sticking out in the corners of the cars (lidars, as you can see it spinning) and some big housing on the roof.

I'm sure they're not bothering with costs at the moment, just collecting data and improving their code.

I thought the video was notable because I'd read elsewhere that Waymo cars struggled with left turns against oncoming traffic. Well the Cruise test car seemed to handle those as well as people stopping in the middle of the street or walking out in front of the car pretty well.

But I'd also read that in the past, they cooked their demos and it really struggled with making impromptu adjustments for unforeseen obstacles or objects intersecting its path.

 

[NW-Bound]

Note the dual roof-mounted lidars, plus the two side lidars. There appears to be another lidar on the front grille.

Upon closer look, the thinggies that stick out from the front fenders may not be lidars, but just vision cameras.

Other than Waymo who makes their own sensors, I have not heard of anyone making small short-range lidars to mount on corners of the car. I could be wrong.

 

[HFWR]

Pertinent to this thread?

(Apologies for the nekkid links; on my phone in a motel...)

https://www.cnbc.com/amp/2019/01/30...-self-driving-trucks-developed-by-embark.html

Includes a discussion of AI and autonomous vehicles.

https://www.ted.com/talks/the_ted_i...hat_the_future_holds_next/up-next?language=en