Separate names with a comma.
Discussion in 'Model S: Driving Dynamics' started by enzo, Dec 9, 2016.
On-ramp to off-ramp....
Hah! Good point.
Whenever I drive to my daughter's in Chicago, I find the Illinois Tollway system ridiculously difficult to navigate if you don't have a transponder (being from Canada, I have one but it only works here). I doubt the car will be able to pull way over to the Cash Only lanes and hand the attendant money!
Well, hopefully my car never starts throwing coins at people. But until then Tesla is going to have to 1) add a checkbox in settings that you have an RFID tag, 2) be able to read a variety of toll plaza signs, 3) add the RFID lanes to the fleet learning database, 4) anticipate the appropriate lane and read the signs to confirm, 5) get through the plaza at the appropriate speed.
While fairly involved, doesn't seem that difficult compared to some of the other feats a level 4-5 car must achieve. So I think it will eventually be available for EAP, in fulfillment of the "on-ramp to off-ramp" marketing promise.
I think negotiating a toll booth situation beyond blindly following the car ahead of it is a pretty tall order. IMO, this level of autonomy will not be achievable until roadways are fundamentally redesigned to be more navigable for automated systems. The reality is that this is pretty unlikely to occur. With roads designed as they are now it's very likely that humans will have to take control in complex environments simply because there is nothing to assist automated systems.
I am more interested in how APv2 is going to handle itself in an underground parking garage. For that matter, how is APv2 going to be able to distinguish the difference between one parking lot from another? How far will it go to find a parking space before it gives up? What if the lot is actually full? does it just drive around in circles until something opens up? There are lots of situations where I can't see how an automated system will have trouble adapting. I see AP 2.0 being able to navigate to the extent that mapped GPS navigation can provide information. Beyond that will be very difficult.
I agree about the parking garages. I could see it getting stuck in infinite loops, not realizing certain areas remain unexplored, and getting in the way of other drivers.
But for toll booths, I don't see it as very complex. Reading signs, finding traversable road, changing lanes... these are what Tesla Vision does best. It's not an open-ended problem like parking.
This would highly depend on the nature of the tollway. Since there are many different toll systems in the US alone, that would require specific programming to account for different situations. In my specific example in Northern Virginia there are no lanes to follow unless you are using EZPass. Even then only a few of those lanes are marked as such. The rest are open asphalt up to the booth itself with no clear signage or directions. A person could (mostly) make a decision on how to proceed. I do not see how an autonomous system could successfully navigate. To say nothing about actually paying a toll.
Realistically, tollways would have to evolve to an automated system closer to what you see in Seattle on the 520 bridge. All car license plates are recorded and accounts are billed accordingly. No slowdown for booths, no tag/pass devices mounted in cars.
Then there is the whole HOV/HOT debacle...
At the end of the day, I don't see how high levels of autonomy can realistically be achieved without accompanying legislation and standards applied to roadways. The closest I think we'll see with APv2 is an alert for the driver to take over in known locations where there is a clear gap in capability.
Like AP 1 was supposed to do?
I wasn't around then so I can't comment. But yes, if true then that obligates them even moreso to make good on the promise this time. And Tesla certainly knows this. And we know it's not a dishonest company, just an overzealous one.