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.
I don't count Tesla as.a player as of today. No one will do business with Elon, tbh, unless they can prove it works beforehand.

Yes, that is why I said that no OEM will license current FSD because liability is too high. OEMs need assurances of safety. They are not going to license a vision-only system just on Elon's word that it is safer than humans.
 
Last edited by a moderator:
  • Like
  • Informative
Reactions: DanCar and flutas

I wonder why he thinks that cameras and radar will be sufficient in the future if they're not now. Is he expecting an improvement in the software to process camera inputs, an improvement in the quality of radar hardware, or some sort of improvement in sensor fusion?

That would also be a massive-blow to innovation if regulators require a certain piece of hardware.
 
  • Like
Reactions: GSP and JB47394
I wonder why he thinks that cameras and radar will be sufficient in the future if they're not now. Is he expecting an improvement in the software to process camera inputs, an improvement in the quality of radar hardware, or some sort of improvement in sensor fusion?

That would also be a massive-blow to innovation if regulators require a certain piece of hardware.
Progress in computer vision basically and sensor tech in general.

I personally think it will be a while, if ever. Too many weird failure modes (an image of a road on a truck etc). I think Lidar will drop in price faster than there is progress in computer vision, now that more OEM:s (esp in China) have them in their models.

 
Last edited:
  • Like
Reactions: Goose66
I wonder why he thinks that cameras and radar will be sufficient in the future if they're not now. Is he expecting an improvement in the software to process camera inputs, an improvement in the quality of radar hardware, or some sort of improvement in sensor fusion?

Yes, I think he is assuming that there will be improvements to the software both for camera vision and radar that will make camera+radar perception good enough to support L3+ autonomous driving. And it is a logical assumption because computer software is improving all the time. We have already seen massive improvements in computer vision and radar in recent years. Years ago, computer vision could only do basic object detection and required stereo vision to do any type of distance measurements. Now a single camera can extract all the information needed for driving (occupancy networks, road geometry, lanes, road signs, hand gestures, traffic lights, curbs, crosswalks, pedestrians, cyclists, vehicles, trucks, etc...) as well as do distance and velocity measurements with decent accuracy. Likewise, we have seen huge improvements in radar. Radar used to be very low resolution. It could be used for basic cruise control to maintain a safe distance from a lead car but that was about it. And low res radar cannot detect static objects so it can cause collisions with stopped vehicles on the road. But now, we have imaging radar that is very high resolution and can accurately measure position and velocity of both stopped and moving objects, and separate between objects close to each other, like distinguishing a pedestrian standing next to a stop sign. It stands to reason that with camera vision that can do all perception on its own and high res radar that can accurately detect and measure position and velocity of static and moving objects in great detail, that with further improvements to reliability, camera + radar will be sufficient for L3+.

That would also be a massive-blow to innovation if regulators require a certain piece of hardware.

Yeah, I don't think that regulators should mandate a certain tech. Regulations should be tech agnostic. They should set safety standards and let manufacturers use whatever tech they want if they can achieve those safety standards.
 
25 States in the US now have laws that support AVs:

State policymakers are seizing the opportunity of AVs and continue to support regulatory frameworks that promote safety and innovation. In April, Kentucky became the twenty-fifth state in our union to have an AV deployment statute. It is a milestone worth celebrating that the United States now welcomes autonomous vehicles in one half of its states, with others possibly joining soon. That is 193 million Americans, representing 56% of the population.

AV statutes generally take action in several key areas. The laws authorize autonomous vehicles - commonly referred to as Level 4 or 5 Autonomous Driving System (ADS) – to travel on the roads of the state. They also contain several safety enhancing elements, such as a requirement that for an AV to be lawfully on the road the vehicle must be able to achieve what is called a “minimum risk condition,” a technology feature that allows the vehicle to safely take itself out of commission in the event of a problem. In addition, state AV statutes appropriately specify that an AV must be capable of complying with all applicable traffic and motor vehicle safety laws of the state. The laws require that proof of financial responsibility for the AV must be submitted to the appropriate state agency and that the vehicle must be properly registered and titled in the state.


That last paragraph is key. The laws require the AV to be able to achieve a minimum risk condition. So Tesla will need to add that feature to FSD before they can think of removing driver supervision.
 
  • Informative
Reactions: GSP
Interesting crash report, and a great example of tiny things, from Zoox. [0]
We discussed that earlier. It is not so tiny; it can cause catastrophic accidents. Manhole covers are prevalent in every major city with underground utility infrastructure.
TJ9CUvz.gif
kGBanI7.gif