-
Want to remove ads? Register an account and login to see fewer ads, and become a Supporting Member to remove almost all ads.
Will the Model 3 become the standard fleet purchase for Taxi Coys.?
- Thread starter Nuclear Fusion
- Start date
-
- Tags
- Model 3
garsh
Re Member
If Tesla pulls off L5 autonomy with the AP2 hardware, I think they absolutely will become standard fleet for a lot of taxi companies. I disagree that the limited cargo space will negatively affect the potential of this. In fact, I think the emphasis on maximum passenger space (at the expense of cargo space) was a deliberate decision based on the future use in the Tesla Network. So many more people these days use Uber and Lyft for day-to-day purposes without any need for luggage space.
BluestarE3
Active Member
TIL the abbreviation for "companies". Thanks!
Then again, coy is an "abbreviation for company in the UK and some Commonwealth militaries".
It's questionable whether or not they have hardware necessary to actually be full-time, fully autonomous, regardless of how good the software is. All depends on how one defines the requirements for fully autonomous driving. Turns out that "L5" has an extremely narrow scope and mainly deals with safety considerations - not causing accidents, rather than strategies for intelligently negotiating various road conditions, communicating with humans, and parking (every day stuff). With narrow requirements, everything is sufficient - with broader, real-world, requirements those sensors will wind up being inadequate.
garsh
Re Member
The engineers at Tesla believe that they have all of the necessary hardware. You can question it, but I see no reason to. We've already seen a demonstration video of a Tesla driving itself.
Again, "L5" is defined in such a narrow manner that the present hardware *could* comply with it and still need human takover for everyday driving communication, cooperation, strategies and tasks. How "fully autonomous" is that? Hardware capability has nothing to do with what they "believe" when practicality is the consideration rather than limited standards compliance. A car driving itself in certain conditions is light years different than a car driving itself in the various conditions handled easily by drivers. So "full autonomy" can mean "limited full autonomy" or "part-time autonomy" - because the term derives from the L5 definition which has (purposefully) narrowly defined capability requirements. Get it now?
L5 is focused on accident avoidance when driver is not at the wheel. Driving goals are pursued in a tactical, not a strategic, manner all the while limited by AI and hardware well short of what any human can easily handle.
garsh
Re Member
I don't know where you're looking for your definitions, but that's simply untrue.
SAE Automated Driving Definitions
Simply nope. Level 5 doesn't include things like finding appropriate parking or dealing strategically with the everyday conditions I've mentioned. The camera's resolution is insufficient for planning a new route due to incident plainly visible only a few blocks down the street, camera rez cant make use of convex mirrors for blind driveways (radar coverage is inadequate) the car would not know to avoid heat sources (car, bldg, road fire), it can't determine how deep water is for safe crossing, it can't listen to or follow instructions from road workers, etc. All of these are hardware limitations and trivial for a 16-yr old human to do.
garsh
Re Member
That's exactly what it includes. That's why it's referred to as "full autonomy".
The camera and radar can detect obstructions. Rerouting is handled by software.
Neither can my eyeballs, but through fleet learning the cars will know that a blind driveway exists there and take whatever precautions are required.
Cameras can detect flames & avoid them, easily.
Deaf people can't listen either. And I already posted a link showing that Waymo cars can interpret hand signals.
Your strawman arguments are all incorrect. The engineers are thinking about all of these scenarios, and more.
Sure, from a limited distance about 10x short of what a human can detect.
Re-routing is subject to map info or fleet info. Route change planning often requires prior knowledge and fails if you happen to be the first one to arrive at incident or an insufficient number of cars aren't updating the correct info. Again, trivial for a human to figure out but not possible for a car without knowing ahead of time there is a problem ahead. The cameras simply can't see that far to make out anything. I don't think you quite get the humongous difference in resolving power between a person and an AP2.0 camera system.
You are actually saying that you can't use a convex mirror? Like in a parking garage or on a telephone poll or end of a home driveway? There is no "fleet learning" when entering a street from a home blind driveway. Even if there was, the car would not safely take the road without all other cars also being autonomous. Think about it.
Yet, heat frequently happens without visible indication of flame. Even if there was a visible indication, the flames could simply be elevated wrt to camera viewing angle - say at the 2nd or the 3rd floor of a large, city block building. Cameras wouldn't see it.
in the real world hand signals or gestures are completely non standard and vague. A person, however, can figure out meaning from context - by following modifications to directions due to other driver behavior. You could potentially force the authorities to standardize their signals.
Of course they are, and improving the autonomous operation will inevitably require hardware upgrades, which was the whole point. It all depends on how you define fully autonomous.
