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

Can level 5 Autonomy be achieved with Hardware suite 2.0?

This site may earn commission on affiliate links.
You can't filter out what LiDAR can't penetrate... ie water.

Driverless cars have a new way to navigate in rain or snow

May be "filter" is a wrong word.

The algorithm differentiates patterns of bouncing back rays coming from a snow flake or a drop of rain versus an obstacle.

Specialty LIDAR does see through water to determine water depths and 3-d map an ocean floor:

"Bathymetric lidar is used to determine water depth by measuring the time delay between the transmission of a pulse and its return signal."

SW_Lidar_image.jpg
 
Last edited:
Neural net produces object recognition. It classifies tractor trailer as tractor trailer and knows, that don't try to drive trough a tractor trailer. It needs only one camera for that. As human with one eye can identify tractor trailer as one sees it.

Properly working neural net should also identify plastig bag or newspaper page floating in the wind as something that you can drive trough if necessary.

Without object recognition a self driving car is impossible to create.

Radar can be used to detect density as well as whether an object is metallic. That's the beauty of it. It can determine a threat regardless of the shape. Elon used a UFO in the road as an example to make this point.
 
Driverless cars have a new way to navigate in rain or snow

May be "filter" is a wrong word.

The algorithm differentiates patterns of bouncing back rays coming from a snow flake or a drop of rain versus an obstacle.

Specialty LIDAR does see through water to determine water depths and 3-d map an ocean floor:

"Bathymetric lidar is used to determine water depth by measuring the time delay between the transmission of a pulse and its return signal."

SW_Lidar_image.jpg
From your link:

"a lower frequency infrared pulse is reflected off the sea surface, while a higher frequency green laser penetrates through the water column and reflects off the bottom."

from the other link:
"He wouldn’t say what depth of snow the cars have successfully navigated. "

he says later:
"Let’s say I could drive at 80 miles per hour in the fog because I had some magical sensor that could see through it"
Implying that his system was incapable of that... whereas radar is.
 
  • Informative
Reactions: Tam
Please go to the NVIDIA YouTube web site and view the presentations they have there
regarding the DRIVE PX 2 computer platform that Tesla is now using in all of its cars.
I don't think anyone is worried about that... as a matter of fact not a single person or news article has questioned Drive PX 2 as far as I know. The thing people question is radar and camera placement.

Please (everyone) stop quoting the nvidia stuff... it has nothing to do with the sensor hardware.
 
Last edited:
  • Like
Reactions: Matias
Elon suggests that we don't bet against the AP 2.0 hardware suite....:)
Tesla dictates what capability is deemed sufficient. For the consumer, the question will be is there a better alternative. If not, then what are you betting against? Being first to market with something a few years ahead of the competition gives them a monopoly. What would be the reason to "bet against", unless it didn't offer what was promised. There are no explicit promises, only very strong inferences as far as what the customer should expect. As long as they don't explicitly promise something they can't deliver, they will be fine.
 
I had a totally different conception of what a fully autonomous car should be capable of vs what the industry has already decided is sufficient. I now can easily see that Tesla can deliver a "Level 5" car in a couple of years.
Yeah, level 5 isn't as autonomous as people think. it's merely "self-driving". It's not required to handle some unforeseen crazy contrived conditions that most people try to pit it against.

"What does it do at 60 mph when clowns attack it from three sides while a child is jumping out in front of it and there's a sand storm?" huh? huh?
 
Yeah, level 5 isn't as autonomous as people think. it's merely "self-driving". It's not required to handle some unforeseen crazy contrived conditions that most people try to pit it against.

"What does it do at 60 mph when clowns attack it from three sides while a child is jumping out in front of it and there's a sand storm?" huh? huh?

The problem will be not so much due to the car's inability to handle "difficult" situations, but rather not handling the situations that occur only 5-10% of the time, but nonetheless will always occur. The consequences of a car hampering traffic flow will be anywhere from a nuisance (can move around) to disastrous (blocking). It does seem that the industry recognized that although such problems would be difficult to solve, the value will mostly outweigh the problems. Better to ask forgiveness than permission.

Unfortunately society will have to accommodate these limitations in some manner until the AI eventually improves. Kind of begs the question of whether or not manufacturers with better AI will be forced to share their technology with their competitors in the name of the public good.
 
The description for Enhanced Autopilot in the design studio says "automatically change lanes without requiring driver input". I'm curious how it will do that on the autobahn with 100 meters rear visibility. For example, Daimler's system has a rear radar with 200 meters range. If the Tesla moves in front of another car that's approaching at much higher speed, how is the other going to slow down in time? I was collecting some material to start a new topic but I changed my mind. Let me add those images and video here:

You can start watching at 0:15. It shows a few cars passing by on the autobahn:

TxTLVzW.gif



image source: Tesla.com
5BxAr2F.gif


image source: Daimler.com
BwLJJCE.gif
 
  • Like
Reactions: Matias
The description for Enhanced Autopilot in the design studio says "automatically change lanes without requiring driver input". I'm curious how it will do that on the autobahn with 100 meters rear visibility. For example, Daimler's system has a rear radar with 200 meters range. If the Tesla moves in front of another car that's approaching at much higher speed, how is the other going to slow down in time? I was collecting some material to start a new topic but I changed my mind. Let me add those images and video here:

You can start watching at 0:15. It shows a few cars passing by on the autobahn:

TxTLVzW.gif



image source: Tesla.com
5BxAr2F.gif


image source: Daimler.com
BwLJJCE.gif
There are very few cars that go 200 mph for starters... second, in autonomous mode it probably wouldn't pass on its own. Third, both cars are moving so your stopping distance (besides subtracting the thinking time) are not accurate.

It's more likely you are driving manually at that point and it'd be the same risk there is now, except with 100m warning giving you and the car behind you around two full seconds to react depending on your speed.

Fourth, if both cars have radar, AEB and TACC then the fast car will automatically adjust.

Fifth, unavoidably neural networks must be customized per location so this possibility is likely going to be accounted for especially in Germany.

I firmly believe in real life it's going to be a non-issue.
 
So let's say a Tesla with Enhanced Autopilot activated is driving behind a slower vehicle on the autobahn at 100 km/h (62 mph) and is about to change lanes. There is another car in the left lane a little more than 100 meters away approaching at 230 km/h (143 mph). Enhanced Autopilot wouldn't see the other car because it can see up to 100 meters behind. Also, according to the design studio, Enhanced Autopilot will do lane changes without any driver input. The speed difference is 130 km/h (81 mph). How long braking distance does the other need to slow down? It seems more than 100 meters would be needed.
 
Last edited:
  • Like
Reactions: Matias
So let's say a Tesla with Enhanced Autopilot active is driving behind a slower vehicle on the autobahn at 100 km/h (62 mph) and is about to change lanes. There is another car in the left lane a little more than 100 meters away approaching at 230 km/h (143 mph). Enhanced Autopilot wouldn't see the other car because it can see up to 100 meters behind. Also, according to the design studio, Enhanced Autopilot will do lane changes without any driver input. The speed difference is 130 km/h (81 mph). How long braking distance does the other need to slow down? It seems more than 100 meters would be needed.

No, autopilot and enhanced autopilot require the user to initiate the lane change via signal... and then they change lanes without additional user input.

Now, for your hypothetical again you are assuming the Tesla is stopped. 143 mph - 62 mph = 81 mph = 36.2 m/s this gives you and the driver behind 2.76 seconds to react.

The recommended speed is 81 mph so even if the Tesla was doing 81 mph you'd have over 3.5 seconds.

Keep in mind German drivers are supposed to be really good so it'd be some foreigner that puts themselves in this situation. EAP is not self driving, the human must be alert and in control. Also keep in mind that there are speed limits on much of the autobahn, it's not all a free for all.

If someone switches lanes in the unrestricted sections without checking their mirrors regardless of additional safety features then they are a complete moron and shouldn't be driving in Germany. It's really that simple.

Furthermore, I've never been to Germany but as I understand it those vehicles going over the recommended speed in unrestricted zones are liable for damages in any crash they are involved in regardless of fault. So it's in the rear driver's best interest to slow down if he see's such an event unfolding.
 
The description for Enhanced Autopilot in the design studio says "automatically change lanes without requiring driver input". I'm curious how it will do that on the autobahn with 100 meters rear visibility. For example, Daimler's system has a rear radar with 200 meters range. If the Tesla moves in front of another car that's approaching at much higher speed, how is the other going to slow down in time? I was collecting some material to start a new topic but I changed my mind. Let me add those images and video here:

You can start watching at 0:15. It shows a few cars passing by on the autobahn:

TxTLVzW.gif



image source: Tesla.com
5BxAr2F.gif


image source: Daimler.com
BwLJJCE.gif
Looking at the pictures, the rear radar has a much smaller angle of view than the cameras. Can it even see one lane over with enough accuracy to detect the car 200m away?