Every morning I use autopilot to get to work, mostly freeways and about 8 miles of it is into the rising sun. "autopilot" instantly informs me the camera is blinded and to take over immediately. Seems like a pretty big flaw if you are really going to claim "Full self driving" Will this get fixed or will it it still be a prepare to take over at any moment type deal
-
Want to remove ads? Register an account and login to see fewer ads, and become a Supporting Member to remove almost all ads.
So How IS FSD Going to work?
- Thread starter Altes
- Start date
UncleCreepy
Member
With the current hardware there is no way that autonomous driving will ever work. Hell, they can't even get the speed limits straight. I'd love to see my car back out of the driveway without running over someone who's in a blind spot. Street covered in snow? No camera can look through this. Street signs blocked by parked vehicles, blinded, blocked or fogged cameras, the list goes on and on. Right now AP is a nice thing to have as long as the conditions are right.
More importantly, I believe that the whole approach is inherently wrong. Most of the time the driver is the problem, and yet they're trying to clone a digital version of the driver. Ever tried to automatically change lanes with FSD? The computer keeps hesitating if there is a car in the target lane. Even if the other driver is clearly (to me anyway) waiting, the car has no clue and keeps wondering if it can safely change lanes.
What it takes is that all traffic is connected so that in this example, the car wanting to change lanes sends a request and cars that are potentially in the way acknowledge the request, thus allowing the car to change lanes. Also, to cover for roads that are covered in snow or otherwise have no markings (road construction) there needs to be a beacon signal coming from the middle of each lane. Again, just an example, there are more things that need to be done before this can ever work.
The biggest problem is that all traffic needs to be networked, and here I don't know how to do that. First off, it can't work properly as long as there are still regular and therefore unpredictable vehicles in the street. They would mess up the whole system. Also, pedestrians/bicycles/Motorcycles obviously can't be included, so how is that going to work? The car would still have to do a lot of guesswork since non-networked vehicles or pedestrians are not predictable.
Yet another huge problem is the liability. If an autonomously driving car causes an accident, whose fault is it? I'm talking about fully autonomous driving where there are no pedals and steering wheel.
Even what Elon calls FSD is not anywhere near a working system. AP can't make a tight turn, drive through roundabouts or slow down before it reaches a lower speed limit. Without intervention, my car would just stay at 80 km/h for a bit when it reaches a 50 km/h sign. AP is great and all and I use it extensively, but it needs a lot of supervision.
CAD $9,200 my ass. I'm not going to pay for something that I believe will never fully work.
More importantly, I believe that the whole approach is inherently wrong. Most of the time the driver is the problem, and yet they're trying to clone a digital version of the driver. Ever tried to automatically change lanes with FSD? The computer keeps hesitating if there is a car in the target lane. Even if the other driver is clearly (to me anyway) waiting, the car has no clue and keeps wondering if it can safely change lanes.
What it takes is that all traffic is connected so that in this example, the car wanting to change lanes sends a request and cars that are potentially in the way acknowledge the request, thus allowing the car to change lanes. Also, to cover for roads that are covered in snow or otherwise have no markings (road construction) there needs to be a beacon signal coming from the middle of each lane. Again, just an example, there are more things that need to be done before this can ever work.
The biggest problem is that all traffic needs to be networked, and here I don't know how to do that. First off, it can't work properly as long as there are still regular and therefore unpredictable vehicles in the street. They would mess up the whole system. Also, pedestrians/bicycles/Motorcycles obviously can't be included, so how is that going to work? The car would still have to do a lot of guesswork since non-networked vehicles or pedestrians are not predictable.
Yet another huge problem is the liability. If an autonomously driving car causes an accident, whose fault is it? I'm talking about fully autonomous driving where there are no pedals and steering wheel.
Even what Elon calls FSD is not anywhere near a working system. AP can't make a tight turn, drive through roundabouts or slow down before it reaches a lower speed limit. Without intervention, my car would just stay at 80 km/h for a bit when it reaches a 50 km/h sign. AP is great and all and I use it extensively, but it needs a lot of supervision.
CAD $9,200 my ass. I'm not going to pay for something that I believe will never fully work.
Tam
Well-Known Member
It's possible that Tesla FSD will be designed to work in perfect road conditions.
One of the complaints about Autopilot is it only takes easy jobs such as straight clear highway roads and leaves hard jobs to human: winding rainy roads as well as city roads...
i have no confidence FSD will come out until atleast 3-5 more years. what you are paying for is just the option to beta test things along the way to that journey. so we're talking maybe level 3 autonomy level 4 autonomy then 4.5... then 4.5.5.. then for 4.5.5.5.5.5.5.5 revision 100.
Similar threads
