Not really but nice try. Supercruise is vaporware until its released and we have no idea how it actually works other than propaganda. Based on that my Tesla is a level 5 autonomous vehicle. Yea right.
-
Want to remove ads? Register an account and login to see fewer ads, and become a Supporting Member to remove almost all ads.
Scary Experience with AP2.0 Autosteer on the Highway
- Thread starter VeryOrange
- Start date
-
- Tags
- Driving Dynamics Model S
Not really but nice try. Supercruise is vaporware until its released and we have no idea how it actually works other than propaganda. Based on that my Tesla is a level 5 autonomous vehicle. Yea right.
Not in my experience in and around Chicago. I get the several straight lines = a curve effect (I call ping ponging but its actually gotten better over time). My Easter trip across the country was much better and I used AS at 70 or 75 without issue for 750 miles each way). I think it depends on the turn and the road for some reason.
Sawyer8888
Member
Exactly. The tension that is required on the steering wheel to disengage AS with AP2 is usually much greater than AP1. This creates the situation of having to apply more force leading to over-steer when attempting to disengage the system.
You have to learn to use the brake while disangaging instead of just grabbing the wheel. AP1 is much more user-friendly. AP1 also doesn't go berserk from time to time with lights and sounds the same way AP2 does.
I have found it easier to simply flick the AP/TACC stalk on the steering column to disengage. But either way works. Fighting the wheel is certainly not the best approach.
In an emergency, I'd rather stick with tapping the brakes. Finding the AP stalk and remembering to push it away (when I usually only pull it towards, up or down) is probably too much to ask (for me, clearly to each their own). I had maybe .5 seconds to take over and avoid crashing into a concrete barrier while going 80mph. Turning the wheel is the instinctive thing to do (for me). Braking hard isn't always the best option (and I'm certainly not used to tapping the brakes).
I'm getting used to it but AP2 doesn't jive with traditional driving habits. It requires that you relearn basic techniques which isn't ideal and leads to these kinds of situations. I am used to fishtailing (ice/snow all over) so I know how to turn and not turn as it requires. I never deviated into a neighboring lane and it was over with after 3 wobbles (didn't actually come closer than 2-4 feet from barrier but that's plenty close for me).
Joe F
Disruption is hard.
My issue with Tesla is their canned responses to reporting problems and trying to get answers as to how AP2 is supposed to work, given the lack of proper documentation.
I asked via mail the other day about how to best report these false positives and general AP2 anomalies. Not happy with the canned reply of "voice feedback to do a bug report". As they state, these are periodically reviewed. Really? If it's critical to the point I'm reporting it, that does not warrant a periodic review. Kind of makes me feel it may never be looked into. Then there's the "mail us the details with time stamps etc" response, and as others have found, the reply is generally "no problems found."
My real issue, and I'm trying again to get Tesla to state definitively, is why do we have to report these problems with AP2 when they state their neural network, fleet learning AI is supposed to "fix" these problems.
Why after a near accident, or even a less severe incident, are we as drivers expected to take our attention off of the road and use a poor system that hardly gets speech to text correct, to report something that should be uploaded automatically once on WiFi?
Sorry, rant mode off.
Kant.Ing
Member
CR or some car magazines (or PC Magazine!) should start reviewing and benchmarking these driver assist features' performance, including software and user interfaces.
AP1 was reviewed. It got stellar marks. Semi-Autonomous Cars Compared! Tesla Model S vs. BMW 750i, Infiniti Q50S, and Mercedes-Benz S65 AMG - Feature
The Best Driver-Assist Cars of 2017
Kant.Ing
Member
Thanks. Exactly, thought I have seen it but didn't read it through because I know the answer.
I think it's a pretty sure bet that GM's geo-mapping which restricts their system to only work in areas it was designed for will be more effective than Tesla's system. AP relies on the driver to know where the system will work or not work properly. That hasn't ended well in many occasions.
In my opinion, a system that automatically restricts it's use in areas it was not designed for is a better design than relying on a typical driver to make those decisions.
You really have to "Disagree" my post when you could actually write out a real response? For the record, Tesla also limits the system to where they think it works best (geofenced) but sh!t happens. GM's system hasn't been exposed to real life conditions, hence my post. Your "rebuttal" does nothing to address this reality vs. fantasy distinction. Remember, on paper anything is awesome.
I wasn't aware of that. Are you sure?
I thought EAP could be activated within the allowed speed range anywhere that the system can adequately detect lane markings.
Tam
Well-Known Member
Current AP2 is a work in progress.
Other companies would not release a feature that was capped at 35 mph on freeway.
But Tesla did.
There's no need to report to Tesla that I almost got rear-ended and killed because Autopilot was capped at 35 mph on freeway while the traffic was 80 mph.
That is not a bug! The system performed exactly what it was intended to do.
Tesla engineers were working hard on how to increase the cap to 45 mph on freeway so we should not give them a hard time and slow them down because their goal was not just 45 mph but eventually 80 mph in the first place!
If you want to test unfinished AP2 out then go ahead and do it.
Otherwise, you can repeat what Volvo complained:
“Every time I drive (Autopilot), I’m convinced it’s trying to kill me,”
Thus, for those who would be comfortable with a finished product, they should wait and abstain from using this current unfinished product.
Do you have skin in the game or not?
That's the nature of current AP2. It works as intended.
Last edited:
Tam
Well-Known Member
AP2 does geofence to control autosteer speed limit.
Currently, surface street Autosteer is 35 mph even though around in my area eve when posted speed limits are as high as 40 to 55 mph.
For surface streets with no lanes at all, Autosteer (maximum at 35mph) works for me too.
Tam
Well-Known Member
Google says:
-------------------
ge·o·fenc·ing
ˈjēōˌfensiNG/
noun
- the use of GPS or RFID technology to create a virtual geographic boundary, enabling software to trigger a response when a mobile device enters or leaves a particular area.
"geofencing turns the lights off when you leave and on when you get home"
-------------------
So as soon as I get off a surface street to freeway, my autosteer would allow me to increase the cap to 80 mph.
It does the opposite as soon as I exit freeway to surface streets, my autosteer would automatically decrease the cap to 35 mph.
jeffro01
Active Member
At some point some of you are going to have to stop taking every new post by every new member at face value... There are so many things wrong with the original post that screams BS...
Jeff
Jeff
jeffro01
Active Member
AP2 is definitely geofenced. I drive past the Highland Park Tesla. Its on IL Route 41. At that stretch it is a legit highway (exit ramp 41 Central Ave is the exit for the Tesla SvC and SC and store). It does not allow AS at highway speeds until Lake-Cook Road, as IL 41 turns to I-90/94. The steering wheel icon lights up as I pass the exit sign. It always comes on with 100 feet of that sign. If I drive in the other direction, I will lose the ability to auto lane change at the same location (but it still allow AS because it was active, if I deactivate it I cannot reactivate it at highway speeds).
Losing auto-lane change at the spot proves the geofence works because it allows higher speed AS but I lose that part of the programming.
Losing auto-lane change at the spot proves the geofence works because it allows higher speed AS but I lose that part of the programming.
