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

FSD with current great? Not possible.

This site may earn commission on affiliate links.
Can you point to where this is already happening? I hope it happens soon, but haven't seen any indication of this in the Charlotte or Atlanta areas...

Get roads ready for self-driving cars, feds say

The federal government is officially telling states and cities: Get your roads ready for self-driving cars.

An updated version of the manual that dictates standards for road signs, traffic signals and pavement markings — the “Manual on Uniform Traffic Control Devices for Streets and Highways,” or MUTA — is in the works, according to the Federal Highway Administration.

The new version follows and incorporates a recently released guidance for self-driving cars called “Preparing for the Future of Transportation: Automated Vehicles 3.0.”
 
If you study the cameras and sensors on current teslas it would appear that Fsd is not possible. Think about the car stopping then having to pull out onto a highway where the speed is excess of 55 mph, the car cannot see to the left and right far enough to make the decision and time the maneuver. Something else needs to come into play for that one task to occur. Watch the camera recordings from all cameras you will understand. The radar and side sensors cannot detect that far.

Like others said if won't be worse than human to handle situations like this. That being said if any car can handle this situation it would be a Tesla, with either FSD or human driver. ;)
 
If you study the cameras and sensors on current teslas it would appear that Fsd is not possible. Think about the car stopping then having to pull out onto a highway where the speed is excess of 55 mph, the car cannot see to the left and right far enough to make the decision and time the maneuver. Something else needs to come into play for that one task to occur. Watch the camera recordings from all cameras you will understand. The radar and side sensors cannot detect that far.

If you look at the diagram of sensors on the HW2 car, there is plenty of overlap and enough range. The radar and cameras should have no problem with your scenario:

upload_2019-3-29_17-4-1-png.391441
 
If you look at the video from all 8 cameras in these scenarios there are blind areas, also the cameras are lower than a human eye so that compromises the sight distance somewhat depending on the hill, grade, and other obstacles blocking views. Watching this has convinced me it cannot safely happen with the current configuration....my opinion...worth some discussion and other inputs.
 
  • Like
Reactions: DanCar
[
If you look at the video from all 8 cameras in these scenarios there are blind areas, also the cameras are lower than a human eye so that compromises the sight distance somewhat depending on the hill, grade, and other obstacles blocking views. Watching this has convinced me it cannot safely happen with the current configuration....my opinion...worth some discussion and other inputs.
All of the (5) forward looking cams including the (2) side forward looking cams are at or above eye level
 
  • Like
Reactions: StealthP3D
If you do the math then I agree FSD can never work with current cameras in all situations.

Consider a T intersection where you are on the side road waiting to join a straight highway.

The car would have to rely on the side cameras looking left and right which have 80m of range. If traffic on the highway is doing 100km/h this translates to cars travelling at 28m/s.

This means that your Tesla has to pull out onto the highway and accelerate to 100km/h (so as not to cause a crash) in under 2.85seconds (80/28)

Teslas are fast but not that fast.
 
Last edited:
If you do the math then I agree FSD can never work with current cameras in all situations.

Consider a T intersection where you are on the side road waiting to join a straight highway.

The car would have to rely on the side cameras looking left and right which have 80m of range. If traffic on the highway is doing 100km/h this translates to cars travelling at 28m/s.

This means that your Tesla has to pull out onto the highway and accelerate to 100km/h (so as not to cause a crash) in under 2.85seconds (80/28)

Teslas are fast but not that fast.
You also need to take into account the car would be traveling away from the oncoming car during the acceleration. So after that 2.85 seconds the car is 50m away from the point it entered the highway.
 
You also need to take into account the car would be traveling away from the oncoming car during the acceleration. So after that 2.85 seconds the car is 50m away from the point it entered the highway.
I simplified it just to show how inadequate the sensors are for situations like this.

Other things to consider.

- That 2.85s includes the turn. How much G force would be acceptable to passengers turning onto the highway. I don't think turning with 100% throttle would be a pleasant experience
- Traction on the turn, what if it has been raining etc.
- State of charge. Performance of Teslas drop at lower battery levels. Can't assume the same acceleration at 10% as 90%.
- What if cars on the highway are speeding. What safety margin. If it's 100km/h do you assume cars are travelling 110km/h, 120km/h. more? All reduce the time you have to react and the speed you need to accelerate to.
- If the camera has a range of 80m, how much latency is there in determining the speed of vehicles approaching. It's a single camera. It doesn't have the benefit of human eyes (stereo with depth perception). Therefore it's going to take some time to determine the speed of the approaching object. All of this comes off the 2.85s.
- With a camera based approach it's going to take longer to work out the speed of smaller objects (motorcycles) as there are much fewer pixels to work with. A motorcycle that's close may have the same signature as a truck that's far away. Get it wrong and you have a Tesla pulling out into the path of a motorcycle going 100km/h....

The depressing thing after mentioning all this and thinking about 'safety margins' is that I'm not even sure the current sensors will be good enough for city T intersections with 80km/h limits. Time will tell I guess.
 
Last edited:
I simplified it just to show how inadequate the sensors are for situations like this.

Other things to consider.

- That 2.85s includes the turn. How much G force would be acceptable to passengers turning onto the highway. I don't think turning with 100% throttle would be a pleasant experience
- Traction on the turn, what if it has been raining etc.
- State of charge. Performance of Teslas drop at lower battery levels. Can't assume the same acceleration at 10% as 90%.
- What if cars on the highway are speeding. What safety margin. If it's 100km/h do you assume cars are travelling 110km/h, 120km/h. more? All reduce the time you have to react and the speed you need to accelerate to.
- If the camera has a range of 80m, how much latency is there in determining the speed of vehicles approaching. It's a single camera. It doesn't have the benefit of human eyes (stereo with depth perception). Therefore it's going to take some time to determine the speed of the approaching object. All of this comes off the 2.85s.
- With a camera based approach it's going to take longer to work out the speed of smaller objects (motorcycles) as there are much fewer pixels to work with. A motorcycle that's close may have the same signature as a truck that's far away. Get it wrong and you have a Tesla pulling out into the path of a motorcycle going 100km/h....

The depressing thing after mentioning all this and thinking about 'safety margins' is that I'm not even sure the current sensors will be good enough for city T intersections with 80km/h limits. Time will tell I guess.

I trust that the Autopilot engineers at Tesla have examined these issues and more. I am sure they did their homework when they first developed HW2, especially since Tesla is making some bold claims about full self-driving capabilities. They must have reasons to believe that the hardware is good enough.
 
I trust that the Autopilot engineers at Tesla have examined these issues and more. I am sure they did their homework when they first developed HW2, especially since Tesla is making some bold claims about full self-driving capabilities. They must have reasons to believe that the hardware is good enough.
There was a time when they thought HW2 was good enough for FSD, then came HW2.5 and HW3.

I'm sure they think HW3 will be good enough, but what are the chances there will be corner cases that they suddenly find aren't handled well enough and need extra processing or sensors to make a foolproof solution. It might work great in daytime but struggle at night if a street light is out. Or maybe it struggles detecting a car with a headlight out. All things that humans naturally adapt to but might require significantly more processing for a computer to handle.

It's a long time before a FSD Tesla is going to be able to drop me off at TBIT @ LAX in a timely manner
 
This means that your Tesla has to pull out onto the highway and accelerate to 100km/h (so as not to cause a crash) in under 2.85seconds (80/28)
Not correct - since as you accelerate you are moving forward, which gives you more than 80 meters. You need to know the reaction time, time it takes to turn to the lane & acceleration to solve this.

BTW, in real world the car coming behind you will slow down a bit if you aren't gaining speed fast enough.
 
Not correct - since as you accelerate you are moving forward, which gives you more than 80 meters. You need to know the reaction time, time it takes to turn to the lane & acceleration to solve this.

BTW, in real world the car coming behind you will slow down a bit if you aren't gaining speed fast enough.

I agree that you will be accelerating and moving forward, but the first part of your acceleration only moves you across the lanes in the road toward the lanes that go in the direction you want to complete your left turn.

Until then, the car will have no/little motion in the direction that actually moves you away from the oncoming traffic. If that takes 2 seconds from a standstill, you have just moved into the lane of oncoming traffic and into an impending collision within fractions of a second.

Also, I don't think you can design an automatic driving system that relies on the cars in the oncoming traffic, some driven by humans, to all brake successfully every time. With enough occurrences, many collisions will result.
 
Last edited:
  • Like
Reactions: sdoorex and DanCar
I looked at the diagram and the "Wide Forward Camera" only has a range of 60 meters.

The "Forward Looking Side Camera" has a 80 meter range, but it appears too narrow to see cars approaching from the side? Hard to tell.
 
I looked at the diagram and the "Wide Forward Camera" only has a range of 60 meters.

The "Forward Looking Side Camera" has a 80 meter range, but it appears too narrow to see cars approaching from the side? Hard to tell.

Here is the diagram again. I added an arc to better show the forward looking side camera. I definitely think it would be able to see cars coming from the side.

upload_2019-5-5_13-42-57.png
 
  • Like
Reactions: DanCar