Let's talk Self Driving Cars again!

Isn't that what Waze does? Not mapping, but for traffic, road hazards, and police presence? Seems to work pretty well.

I am not sure why Tesla dropped the approach. It does not solve all the problems, but makes navigating the busy or more traveled roadways easier.

Of course you and I have been driving without constant updates from the cloud about what lays 1 mile ahead. If a computer needs this help, then who's helping the car that first travels this segment of road at 6 AM in the morning, to warn it that the road had been washed out the night before due to torrential rains?

Perhaps that's why Musk tries to do without help from the cloud, to have the computer being as smart and self-sufficient as a human driver. I agree that this is the ultimate goal, but boy, that's very tough.
 
One thing I have mentioned before, that a human driver often has an easier time than a computer because of the following advantage.

He can leave the car, gets out to look at the road or situation before deciding how to proceed.

A computer can only see from the car. It cannot see around the corner.

Maybe a futuristic SDC can have a drone that it sends out to survey ahead. :LOL:
 
Last edited:
Well some SDC systems only use computer vision.

Reasons are they save costs not having to install things like lidar or ultrasonic sensors for parking.

But also the case that sensor fusion is hard to implement.
 
Yes, it is hard to make sense out of dissimilar sensors because they will not provide directly comparable measurements. Tesla has given up on it.

Yet, Mobileye is intentionally using dissimilar sensors for more robustness. I am impressed. I wonder how far along they have gotten.
 
Well that and I think they're a big vendor of lidars?

Waymo is using lidar.

Not sure what other companies are doing but Tesla may kind of stand out only going with computer vision.
 
Nobody is selling much, because there have been no production cars using lidars. Every Level 4/5 SDC vehicle currently on the road is experimental and they all use lidars to my knowledge.

Waymo has been developing its own lidars for a few years.

I did not know Mobileye also developed its own lidar, but it makes sense that it does as a major SDC developer. Same as Waymo, its goal is to provide car makers with the entire SDC computer and sensor suite.

I don't count Tesla production cars as experimental Level 4/5, because they are Level 2 to the company's own admission. What Tesla has done with just camera vision is quite good, but its performance/reliability is a long way from Level 4/5.
 
Last edited:
Tesla is being sued for fraud on its SDC capability claimed for its car (called FSD for Full Self Driving).

Tesla lawyers admit that it may be a "failure" but not a "fraud", and seek dismissal of the suit.

See: https://www.latimes.com/business/st...ull-self-driving-technology-failure-not-fraud

The FSD fraud suit runs through a litany of claims and promises made by Musk and Tesla about automated technology that will be familiar to anyone who closely follows Musk.

They include a 2016 video that purports to show a Tesla driving itself through the streets of Palo Alto with complete autonomy. Before the video rolls, with the Rolling Stones’ “Paint it Black” as background music, a message reads, “The person in the driver’s seat is only there for legal reasons. He is not driving anything. The car is driving itself.”

Tesla workers later revealed that the video was fabricated, done in multiple takes, with the driving systems failures removed, including a crash into a fence. The video remains on Tesla’s website.

If the case goes forward, it could lead to deposition of Tesla employees who helped develop the technology and reveal what Musk knew and didn’t know about its true capabilities when he made numerous forecasts over the years — including the prediction that there would be a million Tesla robotaxis on the road by the end of 2020, that customers could make $30,000 a year hiring them out, and that their cars would appreciate in value.

Tesla lawyers are attempting to prevent that information from going public.

I remember people ooh'ed and aah'ed watching the demo video. However, I thought there was something fishy there, because real life experience with Tesla AP back then was nothing like that. Too much a performance gap between the impressive "experimental" result and real-life production car. Nope. Fishy.
 
Last edited:
Talking to my sister today she told me about a parking space at work that her car will not let her park in unless she turns off the collision safety features so she has stopped parking there because it is too much trouble.

The parking space is close to the entrance and has a large orange bollard on the front corner to stop large truck drivers turning in from cutting the corner and hitting any car parked there. The car’s software thinks it is a pedestrian standing there and since there is less than a 6” clearance as you turn in, the car won’t allow you drive that close to a pedestrian.
 
^^^ Interesting. The software for object identification needs some improvements.

Still, I would not mind having this safety feature, along with an option to turn it off.
 
.

About human taking over, it depends on whether the car has steering wheel or not. Recall how the SDC enthusiasts talk about removing steering wheel and pedals.

And then, it's awkward for a passenger of a robot taxi to climb onto the driver seat to drive himself. Talk about liability. :)

Heh, heh, it can be done - at least in the movies!

("bad words" warning)

("bad words" warning)

 
Autonomous repossession

One aspect of autonomous driving that never crossed my mind, but it did occur to someone at Ford, is the autonomous repossession of a car when the owner has fallen behind paying the loan. Ford has registered a patent for that process, reported by The Verge here.
Someone at Ford put a lot of thought into all of this.

The patent document describes dozens of ways to remotely and electronically revolutionize the entire repossession process, including liaising directly with lending institutions and police.
 
One aspect of autonomous driving that never crossed my mind, but it did occur to someone at Ford, is the autonomous repossession of a car when the owner has fallen behind paying the loan. Ford has registered a patent for that process, reported by The Verge here.
A friend who is in IT posted that, saying that while OTHER systems get hacked all the time, surely THIS one won't....:facepalm:
 
If you’re in the car when it’s repoed would that be kidnapping?

Yeah, imagine you're on a lonely stretch of road at night and it's 20 below and 10 miles to the nearest house when they shut down your car. People have died in such break-down situations. I'm sure that would be good publicity for Ford.

Now FORD will stand for "Fix Or Repossess Daily.":LOL:

I've also heard that law enforcement is pushing for a similar capability to shut down your car at will. I'm sure that would never be abused, my tag line not withstanding. My BIL pulled over a pretty girl (ostensibly for a traffic infraction) and ended up marrying her (wife 2 of 3.) Lots easier with a kill switch, I would think.

Or you get in your car and your infotainment screen says "Your car has been impounded by the IRS until such time that you pay your remaining balance and penalties - including a $100 fee for disabling your car. Have a nice day." Signed, agent 2023-79,885.
 
What I would design for that system is a governor kicks in, maybe to 20 mph, and a beeping horn, but full steering, brakes etc. This way the person could still control the car.

Of course, if you are on the freeway, this could be problematic too.
 
Here’s a company that is doing some interesting stuff in the field of self driving cars.
This interview explains it a lot better than I can. No guarantees, of course.

Pick episode 19.

https://pod.link/1677810412/episode/03a16397446a3d4f5996a9e5829465f7

Here is the website
https://www.ghostautonomy.com/company
The Ghost Autonomy Engine is a flexible platform that helps leading automakers bring autonomy to market faster. Ghost autonomy engine. Software. AI software powers how Ghost sees, understands, and drives. Hardware. Runs on scalable reference hardware designed for OEMs to customize and package.
 
Last edited:
Here’s a company that is doing some interesting stuff in the field of self driving cars.
This interview explains it a lot better than I can. No guarantees, of course.

Pick episode 19.

https://pod.link/1677810412/episode/03a16397446a3d4f5996a9e5829465f7

Here is the website
https://www.ghostautonomy.com/company


If off-topic, I apologize up front.



My opinion about why self-driving cars will not supplant human drivers anytime soon: We hear of 40,000 people being killed every year by human drivers. We have low expectations of human drivers. We're mostly all drivers here (or were) and we know how a moment's inattention can lead to a frightening close call. So, humans as lousy drivers isn't a big, news story.



So when humans screw up - it's not news.



BUT, let ONE (just one) self driving car screw up and it's front page news on every channel. So, while we don't expect much from our fellow human drivers, we DEMAND perfection from self-driven cars. With that dichotomy, self-driven cars will be a long time coming. YMMV
 
So when humans screw up - it's not news.

BUT, let ONE (just one) self driving car screw up and it's front page news on every channel. So, while we don't expect much from our fellow human drivers, we DEMAND perfection from self-driven cars. With that dichotomy, self-driven cars will be a long time coming. YMMV


While the above is true, we are still a very long way from an SDC beating a good driver, and I mean a good and attentive, not a drunk driver or someone who texts while driving.

If you use a bad driver as the basis for comparison, as a reference, then I don't want anything to do with your SDC. In fact, such an SDC should be banned.

About how hard SDC is, the interested can read the following article in Spectrum, a magazine of IEEE (Institute of Electrical and Electronics Engineers), written by a professor of Computer Science of Duke University, published July 30, 2023.

https://spectrum.ieee.org/self-driving-cars-2662494269
 
Last edited:
I don't know what's going on in Scottsdale, but I see what appears to be a self driving car very time I go there. Don't see any anywhere else.
 
I don't know what's going on in Scottsdale, but I see what appears to be a self driving car very time I go there. Don't see any anywhere else.
One of Waymo's (Google's) main locations for their self-driving program is based in Chandler, AZ. Not surprised you would be seeing their vehicles there.
 
One of Waymo's (Google's) main locations for their self-driving program is based in Chandler, AZ. Not surprised you would be seeing their vehicles there.

Ah makes sense. I would think that Old town Scottsdale with it's plethora of pedestrians crossing streets would present a challenge to self driving cars.
 
Waymo established an operation in Chandler 6 years ago (2017). I don't know how many cars it had, but I immediately saw their cars running all over town, even in my residential neighborhood which is not close to their Chandler garage.

Ever since, every time I made an errand, I ran across one or two Waymo cars: Chandler, Tempe, Mesa, Phoenix, mainly the east side of Metropolitan Phoenix. I don't know the territory of their testing ground, but two years ago when they announced a trial of rides to the public, I went on the map and saw that the geo-fenced area for public ride was quite small. In fact, it did not even extend to my home. These guys wanted to play it safe.

Waymo had a presentation of the original sensor suite, and I was able to correlate that with what I saw back then. Besides the main 360-deg lidar and forward-looking cameras mounted on the top, they used to have a small rotating lidar for the rear looking down at the ground right behind the car, a chin-mounted lidar looking down from the front bumper, lidars at the corners of the car. There were vision cameras too besides the main forward looking cameras, looking at the ground all around the car. There were also radars, ultrasonic sensors all around the car. A heck of a lot.

Over the years, I saw several evolutions of their sensor suites. I happened to be able to capture a photo of the latest Waymo car just 2 days ago. It's significantly different from what I saw initially.

There's still a top lidar on top of the roof dome. Underneath the lidar are 6 roof-mounted cameras to cover 360 degrees. Looks like there are fewer sensors now. What's weird is an inverted corrugated cone mounted on the rear trunk lid, above the license plate. There's a cone like that on the fender. I wonder what that could be.

They kept testing and testing, without much exciting stuff to report. And their sensor suite kept changing. I slowly lost interest, until I captured the posted photo. There was a driver in the car. I could not tell if there were passengers.

Just now went on Waymo Web site. They have extended the operation for public ride hailing to 225 sq.mi. of the Phoenix metropolitan area. And the area includes Old Town Scottsdale now.

Waymo is also doing something in other cities, but as mentioned I have not been following this. But recently, I saw a Cruise car. Yes, the Cruise robotcars that have been testing in SF, starting at about the same time Waymo started here.

10965-albums220-picture2831.jpg
 
Last edited:
It's here folks [emoji16]

Full Commercialization of Robotaxis Arrives in San Francisco

Robotaxi regulator, the California Public Utilities Commission, after hours of testimony from supporters and opponents, voted on Thursday to allow GM’s Cruise and Alphabet’s Waymo to charge riders for driverless robotaxi service, day and night, anywhere in San Francisco, with no cap on fleet sizes. They can now commence full commercialization of robotaxis in San Francisco.

https://wolfstreet.com/2023/08/11/full-commercialization-of-robotaxis-has-arrived-in-san-francisco/
 
Back
Top Bottom