Can level 5 Autonomy be achieved with Hardware suite 2.0?

[azred]

You should sit outside a DMV and watch 16 yr olds take their driving tests

I'll never forget the time many years ago when I was at the Tucson MVD and observed a woman who was clearly struggling with the vision exam. The exasperated examiner told her to look out the front window and telll him what the billboard across the street said. She read it to him with some difficulty -- and was immediately passed. I hope regulatory authorities realize what a low bar AI in our cars should be required to exceed. Any fool can get and keep a driver's license.

I live in AZ and we have had over 1600 wrong way driver incidents in the past year, several resulting in fatalities. Level 5 can't happen soon enough.

 

[brkaus]

Here's a 6 year old paper on the topic
Real Time Speed Estimation of Moving Vehicles from Side View Images from an Uncalibrated Video Camera

Thank you!

 

[Reeler]

Autopark is an extremely constrained subset of autonomous driving.

OK. Why can't Tesla get the car to do it? Answer that and then repeat that answer for all the Koolaide drinkers that think AP 2.0 will do full L5 autonomy.

 

[stopcrazypp]

OK. Why can't Tesla get the car to do it? Answer that and then repeat that answer for all the Koolaide drinkers that think AP 2.0 will do full L5 autonomy.

I don't think necessarily parallel parking is easier than just driving in the city. There are plenty of people who can't parallel park, but drive just fine in the city.

Your margin of error is measured in inches for parallel parking, with restricted sight, and the sensors may not be good enough for that (esp. for AP 1.0 which is what you are discussing, we don't know how AP 2.0 will do yet).

 

[JeffK]

OK. Why can't Tesla get the car to do it? Answer that and then repeat that answer for all the Koolaide drinkers that think AP 2.0 will do full L5 autonomy.

It can do that in the many situations. The times where it fails is where another vehicle is above the ultrasonic sensors. A side camera would solve that issue.

You can't seem to grasp the fact that autonomy does not mean perfection. It's going to keep troubling you until you come to grips with it. How many humans have messed up parking jobs? Is it better than a human on a statistical basis?

I'm sure we can all find a video and say I could have done that better. Yet, the same is true with a video of a human making a parking mistake. We could watch the video and point out all the things the human did incorrectly.

example:

Now here the driver had the ultimate choice of following the advice that the car can autopark there. If the driver chose the space in front of the truck it would have been fine. If the driver would have initiated the autopark a few seconds sooner it might not have happened as he wouldn't have been as far forward. The truck bumper height was just a little too high for the ultrasonics. The driver probably didn't consider this, I know I wouldn't have. It's just a mistake. There's not really blame, stuff happens. It doesn't mean autopark is not capable at all of self parking.

 

[Canuck]

OK. Why can't Tesla get the car to do it? Answer that and then repeat that answer for all the Koolaide drinkers that think AP 2.0 will do full L5 autonomy.

Please drop the Koolaide comments. They only make you look silly and take away from a respectful conversation on the subject.

Tesla has already shown that it can get to level 4:

Of course, this is level 4 only because Tesla is waiting on regulators. Take the driver out of the seat (he wasn't needed anyway) and it's arguably level 5. What I think will cause concern for the regulators is dirt, snow, ice, etc. on the cameras. I think Tesla will need to develop a cleaning system for the cameras, like a stream of washer fluid followed by a burst of air, to get it to level 5. So perhaps you are right that current AP2.0 hardware can't do full L5 but in my view that is some time off in any event. The regulators will want Level 4 to be proven with drivers in the seat for at least a year, and probably more, before L5 is approved and by that time I'm quite certain Tesla will have a cleaning system for the cameras.

 

[Barklikeadog]

Please drop the Koolaide comments. They only make you look silly and take away from a respectful conversation on the subject.

Tesla has already shown that it can get to level 4:


Of course, this is level 4 only because Tesla is waiting on regulators. Take the driver out of the seat (he wasn't needed anyway) and it's arguably level 5. What I think will cause concern for the regulators is dirt, snow, ice, etc. on the cameras. I think Tesla will need to develop a cleaning system for the cameras, like a stream of washer fluid followed by a burst of air, to get it to level 5. So perhaps you are right that current AP2.0 hardware can't do full L5 but in my view that is some time off in any event. The regulators will want Level 4 to be proven with drivers in the seat for at least a year, and probably more, before L5 is approved and by that time I'm quite certain Tesla will have a cleaning system for the cameras.

agree 100%

 

[Daniel in SD]

These old threads are great. It's amazing how little has changed in the last five years.
Hopefully 5 years from now we will finally have good auto park.