Welcome to Tesla Motors Club
Discuss Tesla's Model S, Model 3, Model X, Model Y, Cybertruck, Roadster and More.
Register

Autonomous Car Progress

This site may earn commission on affiliate links.
My argument was this makes sign reading less important as with a digital system the online maps have to be updated in real time. Or they have to use some kind of V2E communication. The speed sign itself is really only for legacy purposes for older vehicles or offline vehicles.

Sure, digital displays and V2E (or V2I) is a useful and important topic in itself. But there are plenty of non-digital ”dynamic” speed signs too. The most common in the world I guess being temporary ones for road work zones and the like but there are also countries that actually swap their physical speed signs a couple times a year due to seasons. Finally, roads evolve and change and non-digital speed signs also change due to this evolution constantly somewhere.

Certainly one day all of this may be replaced by something digital, but that would seem much longer in the future than Tesla’s public robotaxi plans.
 
I don't believe Subaru does with their Eyesight system.

Dissecting Subaru Eyesight Stereo Cameras

Anecdote:

I am by means no expert on all these different systems (I’m most familiar with MobilEye as well as the usual autonomous news) but I also believe there are a couple of different competitors or in-house solutions competing with MobilEye at various makers, while MobilEye is by far the most common.

I could name some names but striving for accuracy I’d have to double check. Anyway, I agree, there are others.
 

I wonder if the intersection caused the AP computer to recognize him as standing in the drivable space. One of the big things I think AP is sorely missing right now is confident path prediction. My car doesn't understand that when another vehicle is slowly turning right off the road, it will be out of our space by the time we get there. But it doesn't understand that and treats the vehicle as a stopped obstacle until it's completely out of the roadway.

Either way, got to love the very clear "What the heck, man?" hand gestures in that clip!
 
My car doesn't understand that when another vehicle is slowly turning right off the road, it will be out of our space by the time we get there. But it doesn't understand that and treats the vehicle as a stopped obstacle until it's completely out of the roadway.

I've noticed that too but it has gotten a bit better over recent updates. I suspect that is something that Tesla will be able to completely fix in future updates.
 
  • Like
Reactions: willow_hiller
https://www.sacbee.com/news/california/article238238929.html
A Silicon Valley startup [Plus.ai] has completed the first coast-to-coast commercial freight trip made by a self-driving truck, according to the company’s press release. Plus.ai announced on Tuesday that its truck traveled from Tulare, California, to Quakertown carrying over 40,000 pounds of Land O’Lakes butter.

And the company says the trek went as smooth as butter.
A self-driving truck delivered butter from CA to PA in 3 days
A Silicon Valley startup has completed what appears to be the first commercial freight cross-country trip by an autonomous truck, which finished a 2,800-mile-run from Tulare, California to Quakertown, Pennsylvania for Land O’Lakes in under three days. The trip was smooth like butter, 40,000 pounds of it.
...
About 10 to 15 companies nationwide are working on autonomous freight delivery, Ives said. That includes San Francisco-based self-driving truck startup Embark Trucks, which last year completed a five-day, 2,400-mile cross-country trip. But that truck carried no freight.
 
It's fun to see the race to autonomy. Of course I think Tesla will win, but what do I know. Waymo seems very smooth, but also in a very controlled environment. If Tesla can get over the hump, they're training their cars for anywhere. I do enjoy being a guinea pig for it though. The cross country truck doesn't give much detail about driver intervention. I can't image there was zero. It does make sense for trucks though...the Tesla is great at just staying between two lines.
 
  • Like
Reactions: APotatoGod
The article in the Mercury news said the following:
" The truck, which traveled on interstates 15 and 70 right before Thanksgiving, had to take scheduled breaks but drove mostly autonomously.
It would appear that there was human intervention based on that statement which means it did NOT drive the entire route autonomously.
Sounds like they may have mapped and pre programmed the route rather than relying entirely on cameras and radar/lidar equipment.
I suspect if Tesla did the same thing they could do it fully autonomously right now.
The mapping method (along with camera & radar) is what GM is using for their super cruise...i.e. in theory, as long as you are on a mapped highway (that is not suddenly under construction) the car/truck will be able to handle the lanes.
 
Last edited:
Chinese robotaxi:


Interesting that they have traffic lights send a signal to the robotaxi telling it when it is red or green. They also have other vehicles communicate with the robotaxis to tell it what they are going to do at intersections. So I wonder if the robotaxi actually does not detect traffic lights at all with its cameras or if it does and the signal from the traffic light is just a safeguard.
 
Chinese robotaxi:


Interesting that they have traffic lights send a signal to the robotaxi telling it when it is red or green. They also have other vehicles communicate with the robotaxis to tell it what they are going to do at intersections. So I wonder if the robotaxi actually does not detect traffic lights at all with its cameras or if it does and the signal from the traffic light is just a safeguard.

I've always thought that V2V or other types of vehicle communication inevitably fails because of the need for near perfect adoption rates. If not every stop light has the standard, then your car needs the ability to read other stoplights; and if you're developing a way to read other stoplights, why bother with the communication standard?

I guess this assumption breaks down in a control-economy like China. If they want, the government can mandate that every stoplight in the country follow a standard, and pour resources into replacing them all rapidly.
 
I've always thought that V2V or other types of vehicle communication inevitably fails because of the need for near perfect adoption rates. If not every stop light has the standard, then your car needs the ability to read other stoplights; and if you're developing a way to read other stoplights, why bother with the communication standard?

I guess this assumption breaks down in a control-economy like China. If they want, the government can mandate that every stoplight in the country follow a standard, and pour resources into replacing them all rapidly.
But you'll always have the case where a defective light doesn't send the proper signal
 
  • Informative
Reactions: willow_hiller
Found this amazing photo on Reddit. A HW3 vehicle identifying a traffic cone despite it being run over and smashed: Extreme Cone Identification

I wonder if the uniformity of traffic cones makes them uniquely easy for Autopilot to identify, or whether this is a sign of things to come for Tesla. Gives me hope for classifying things like stop signs, crosswalks, traffic lights.
 
  • Informative
Reactions: APotatoGod
Found this amazing photo on Reddit. A HW3 vehicle identifying a traffic cone despite it being run over and smashed: Extreme Cone Identification
I wonder if the uniformity of traffic cones makes them uniquely easy for Autopilot to identify, or whether this is a sign of things to come for Tesla. Gives me hope for classifying things like stop signs, crosswalks, traffic lights.
That is pretty good but at the same time it will recognize a kids in an orange shirt as a cone :) -- Tesla Autopilot Confuses Boy In Orange Shirt For A Cone In Brazil