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

Current Tesla hardware is not capable for autonomous city driving says Musk

This site may earn commission on affiliate links.
For context for those who can't (or don't want to) watch, he's saying that sub 10mph is easy because they can stop within range of the ultrasonics. Above 50mph in a freeway environment it's also quite easy because there are less variables. He's saying that the ~15-50mph range in city driving is tough because there's more to look for such as bikes, kids playing, open manhole covers and closed roads.

He also mentions regulators may want 2-3 years after cars are actually better at driving than humans before they'll allow it. He says the cars will essentially run in "shadow mode" so it can provide data to regulators showing where and when the autonomous would have been right or wrong.

There's also a bad joke in there about "premature regulation" :rolleyes:
 
The current car has no way to look left and right at stop signs, etc. I'm not sure if the cone of the front camera is large enough to pick up a typical traffic light.

In addition to the more complex environment that will require a lot more thinking on the car's part, you need new sensors to provide the information the car has to think about.
Walter
 
Will the X initially ship with the hardware required for autonomous city driving? I'll watch the video later, but I would bet the hardest part is deciding whether a lidar is needed, or whether image processing from cameras is sufficient for the city. It could be that both may be needed, so more hardware.

Concerning shadowing, this comment makes sense. The current Google cars driving around record all sensor/camera input for future analysis. Any time something "unusual" is encountered it gets added to the database of scenarios the upgraded software needs to process correctly. Essentially, the cars being driven around provide an ever increasing number of off nominal cases that allow the software to get better and better over time. So if now say 99.999% of unusual conditions are properly processed, when they find 100 more that don't and then modify the code to handle those, you get to 99.9999%. Now it is simply about adding to the trailing number of 9's until you get the code to the level of being as good as the best human driver. And the code updating process is never "done". All accidents that end up happening that could have been prevented will be added to the baseline code after being reviewed. Some accidents are unavoidable.

RT
 
Will the X initially ship with the hardware required for autonomous city driving? I'll watch the video later, but I would bet the hardest part is deciding whether a lidar is needed, or whether image processing from cameras is sufficient for the city. It could be that both may be needed, so more hardware.

Concerning shadowing, this comment makes sense. The current Google cars driving around record all sensor/camera input for future analysis. Any time something "unusual" is encountered it gets added to the database of scenarios the upgraded software needs to process correctly. Essentially, the cars being driven around provide an ever increasing number of off nominal cases that allow the software to get better and better over time. So if now say 99.999% of unusual conditions are properly processed, when they find 100 more that don't and then modify the code to handle those, you get to 99.9999%. Now it is simply about adding to the trailing number of 9's until you get the code to the level of being as good as the best human driver. And the code updating process is never "done". All accidents that end up happening that could have been prevented will be added to the baseline code after being reviewed. Some accidents are unavoidable.

RT

No, it won't. They have to determine what hardware they need, and they won't know that until the vast majority of the software work is done.
 
That's true. But if you take a look at autopilot description in the design page, it's written so, that it may give you wrong impression. What does "drive automatically in rush traffic" mean? Many will understand, that car drives by itself in city.
Disagree. I would say that to most people "rush hour traffic" means going slowly on a freeway or highway and the coming Autopilot feature will surely do that.
 
Better a diamond with a flaw than a pebble without.
The perfect is the enemy of the good.

If we're concerned with premature regulation, we must also be concerned with premature optimization.

I don't want to read about what's not possible -- I want to read about what's possible! I applaud Elon for the aggressive work he's doing, and the progress that we're making as a whole - whether Tesla, Daimler, etc. - at making a better, more automated world. Who cares if my car can't deal with 30 mph city driving for a decade? I don't want that to stop the car from being able to put itself into the garage and connect the charge port. I don't want that to stop the car from being able to handle freeway / controlled-access highway driving.
 
I like the autopilot analogy. Reminds me of an airline pilot, letting it cruise automatically, taking control in complex situations. I'm happy with that. In a way I would prefer to make decisions on when to merge into oncoming traffic, changing lanes in the city etc.

it wasn't so long ago I thought cruise control was a bonus. Then I had smart cruise now on the MS smart cruise plus lane change. It's a small but interesting evolution.

I like driving too much it would be a shame to have a completely automated tesla. It is coming but if I wanted chauffeur mode I'd buy a cheap and uninteresting car.
 
The current car has no way to look left and right at stop signs, etc. I'm not sure if the cone of the front camera is large enough to pick up a typical traffic light.

Hi Walter,

I don't either whether the front camera will pick up a traffic light, but the current system is looking at the break lights of the cars in front.

Larry