But what firmware version are you on? The restrictions are new in the last couple of releases.
-
Want to remove ads? Register an account and login to see fewer ads, and become a Supporting Member to remove almost all ads.
Another AutoPilot Easter Egg NOT!!!!
- Thread starter boonedocks
- Start date
But what firmware version are you on? The restrictions are new in the last couple of releases.
alseTrick
Active Member
Haha. Wut.
Tesla is programming their software to abide by the laws of the road and now some drivers are getting pissed off that they can't speed?
THE HORROR!
This is appalling. Tesla is determined to make AP 1 useless. The hype train for AP 1 is over, they already have your money. Now it's all about CYA.
A5 Looks like a "hwy" that we can also set to any speed under 90mph in the USA. Not apples to apples
SomeJoe7777
Marginally-Known Member
I saw this exact behavior today while driving on the Sam Houston Tollway in Houston.
I was driving Northbound on the West Sam Houston Tollway in Houston, TX, in the leftmost inside lane, traveling 75 MPH with TACC and Autosteer engaged. A police vehicle was stopped on the left shoulder of the road. As I got close to the police vehicle, I received the same message as wk057 showed above -- a yellow warning triangle message on the dashboard that stated "Autosteer restricted to 50 MPH". The vehicle's speed was 75 MPH and began slowing down, the speed limit was 65 MPH, which was also properly registered by the vehicle and was showing on the dashboard.
This is apparently a new safety feature of autosteer. At the time this video was taken, the vehicle is on firmware version 8.0 (2.50.114). I believe the safety feature is to prevent a high differential lane speed between your lane and an adjacent lane when autosteer is engaged. The vehicle saw the stopped car to it's left, and interpreted that the differential speed between my vehicle and the stopped vehicle was too high for autosteer, and so began to limit my speed to 50 MPH.
I have dashcam video of this:
You can look at the silver vehicle on my right and see that my vehicle began to slow relative to it. I disengaged autosteer (you can hear the autosteer disengage tone), and sped back up to my previous cruising speed of 75 MPH, and reengaged autosteer. No restriction was in force upon re-engagement.
I believe Electrek has also posted about this issue.
I have two problems with this:
1. Sudden deceleration without warning is not a good idea. Now, in my case, the deceleration wasn't very sudden. I saw the dashboard warning, realized that the vehicle was slowing down and pressed the accelerator to speed back up. I caught it at about 68 MPH (a deceleration of 7 MPH from my cruising speed). I disengaged AP, sped back up to 75 MPH, and reengaged AP.
2. No documentation or mention of this new behavior. This is the major issue. If I had known about it, I would have expected it coming up on the stopped police vehicle and been ready for it. But without any mention of this in the release notes, it caught me totally by surprise. Not good, Tesla.
I don't know about the laws in Texas, but in Oregon you are required by law to either change out of the lane, or slow down, if there is an emergency vehicle stopped adjacent to your lane. So in that sense to me it seems like AP was doing exactly what it should...
lol. AP1 has no idea what is an emergency vehicle and what is not.
There's no excuse for this behavior. There is no reason whatsoever for autopilot to have been in any way restricted in any of the scenarios where I've had this happen or have seen others report this happen.
Tesla really needs to get their **** together.
Are you sure about that? Given that this is new behavior in the latest release, it is entirely possible that AP now recognizes flashing lights as an emergency vehicle in an adjacent lane and drops speed accordingly. Of course, since we have no release notes or explanation, we are all just theorizing. It would be interesting to see if anyone has noted this exact same behavior under the same conditions (divided/limited access highway) with a vehicle pulled over in the left breakdown lane when you are in the left lane (next to it) and the vehicle has no flashing lights on.
If you live in a town where traffic lights respond to approaching emergency vehicles when the flashing lights are on, you've already seen a demonstration of a means by which an automated system (the traffic light controller) can modify it's behavior based on whether or not it detects an emergency vehicle with it's flashing lights going. Typically these systems are triggered by either an infrared transmission by the emergency vehicle but also by strobe lights. It is certainly not out of the realm of possibility that the AP1.0 forward looking camera has now been trained to respond to flashing lights in the lane ahead of (or adjacent to) the lane the car is in. It could simply be programmed to respond to flashing lights.
Right now, we don't have enough empirical data to know WHY AP1.0 chose to decelerate in the situation you encountered. The fact it gave you a warning is interesting. It certainly seems more LIKELY it was because of the high relative speed differential and might align with the new restrictions posed on non-divided/non-limited access roads (where the change of high relative speed differential is greater), but right now we just don't know....which makes it all the more frustrating for all of us.
Last edited:
Uhhh, no. I don't think normal people realize how dumb autopilot is. There are several small pieces of functionality together that make it seem like something intelligent is going on, but that's just not true. Tesla freaked out over the crashes, and now is making AP worse trying to make sure another one doesn't happen, even if it becomes unusable in the process.
While I'm not sure I agree with the term "freaked out", I do suspect that this is in response to behavior and incidents to date. My point was that none of us "know" for sure and that it is absolutely technically possible for the autopilot software to have been written to now respond to flashing lights it detects by slowing down and/or a high closure rate on a vehicle in front of it or in an adjacent lane by slowing.
- Can the current forward looking camera detect flashing emergency lights (at least the strobe/LED style lights used by many emergency vehicles now and what we saw in the video? Yes.
- Can the software that exists (and we know has changed in the latest update) have a new if/then statement which says "If flashing lights detected, slow to 50 mph"? Yes.
- Can a radar either use doppler or position differential data to determine closure rate? Yes.
- Can the software that exists (and we know has changed in the latest update) have a new if/then statement which says "If closure rate is greater than X, slow to 50 mph"? Yes.
Personally, I don't think the system responded to flashing lights -- far more likely is closure rate above a given threshold (in my opinion and it's worth exactly what you paid for it), but unless any of you are Tesla software developers/senior leaders at Tesla or have rooted your cars and examined every line of code in the latest update, neither you or I "know" for sure.
Call me pedantic, but lots of people on this forum (and in this thread) "know" things or state things as facts when they are simply theories or opinions....I try to distinguish between those, but I'm old fashioned that way. Must be something I learned from my 5th grade science teacher...
FACT: The AP software in the latest release changed.
FACT: There are now limitations on what speed the AP can be set to in different road types.
FACT: The AP is now apparently (a) issuing a warning and (b) slowing to 50mph under certain conditions.
Theory: The AP is now slowing to 50 mph when it sees emergency flashers
Theory: The AP is now slowing to 50 mph when there is a car stopped in an adjacent lane.
Theory: The AP is now slowing to 50 mph when it detects closure rate above a given threshold in the lane in front of it or an adjacent lane.
Theory: Tesla did this because the government told them to.
Theory: Tesla did this because of the crash.
Theory: Tesla did this because of a previously unknown limitation in the AP hardware and/or software.
Theory: Tesla did this because they are concerned about liability.
Theory: Tesla did this because they want to screw over their customers.
Opinion: People at Tesla "freaked out". (define "freaked out)
Opinion: Tesla Autopilot is dumb. (define "dumb")
Opinion: My Tesla sucks with the new AP software. (define "sucks")
Opinion: Tesla screwed me over. (define "screwed me over")
I currently lack the data to prove or disprove any of the theories I just listed. I suspect some are more likely than others, but that is simply an opinion, I lack empirical proof. There are people on planet earth (specifically certain Tesla employees/leaders) who know which of the theories listed above are facts and which are not true. I do not. See the difference?
- Can the current forward looking camera detect flashing emergency lights (at least the strobe/LED style lights used by many emergency vehicles now and what we saw in the video? Yes.
- Can the software that exists (and we know has changed in the latest update) have a new if/then statement which says "If flashing lights detected, slow to 50 mph"? Yes.
- Can a radar either use doppler or position differential data to determine closure rate? Yes.
- Can the software that exists (and we know has changed in the latest update) have a new if/then statement which says "If closure rate is greater than X, slow to 50 mph"? Yes.
Personally, I don't think the system responded to flashing lights -- far more likely is closure rate above a given threshold (in my opinion and it's worth exactly what you paid for it), but unless any of you are Tesla software developers/senior leaders at Tesla or have rooted your cars and examined every line of code in the latest update, neither you or I "know" for sure.
Call me pedantic, but lots of people on this forum (and in this thread) "know" things or state things as facts when they are simply theories or opinions....I try to distinguish between those, but I'm old fashioned that way. Must be something I learned from my 5th grade science teacher...
FACT: The AP software in the latest release changed.
FACT: There are now limitations on what speed the AP can be set to in different road types.
FACT: The AP is now apparently (a) issuing a warning and (b) slowing to 50mph under certain conditions.
Theory: The AP is now slowing to 50 mph when it sees emergency flashers
Theory: The AP is now slowing to 50 mph when there is a car stopped in an adjacent lane.
Theory: The AP is now slowing to 50 mph when it detects closure rate above a given threshold in the lane in front of it or an adjacent lane.
Theory: Tesla did this because the government told them to.
Theory: Tesla did this because of the crash.
Theory: Tesla did this because of a previously unknown limitation in the AP hardware and/or software.
Theory: Tesla did this because they are concerned about liability.
Theory: Tesla did this because they want to screw over their customers.
Opinion: People at Tesla "freaked out". (define "freaked out)
Opinion: Tesla Autopilot is dumb. (define "dumb")
Opinion: My Tesla sucks with the new AP software. (define "sucks")
Opinion: Tesla screwed me over. (define "screwed me over")
I currently lack the data to prove or disprove any of the theories I just listed. I suspect some are more likely than others, but that is simply an opinion, I lack empirical proof. There are people on planet earth (specifically certain Tesla employees/leaders) who know which of the theories listed above are facts and which are not true. I do not. See the difference?
Last edited:
This is the quote of the year. People seem to really overestimate the technical capabilities of AP1. Quite honestly, I think AP1 was released with fewer limitations than it should have been and Tesla is working to slowly reel things back in. When you explain to someone with a basic understanding of mechanics just how dumb AP1 is the response is often awe. Most people think there's this whole suite of sensors doing heavy lifting when it's basically radar cruise control with lane keeping but people see that steering wheel turn itself and instantly assume the car is hiding HAL 9000 in the trunk.
I think if more people understood just how dump AP1 is they really wouldn't want it doing the things they try to make it do. This is why I don't have a lot of sympathy for people that complain about the "hold the wheel" nags on the expressway. The fact that some people don't want to hold the wheel for an extended period of time makes me think there's a lack of understanding about the system's capabilities. And I fully realize that's not a popular opinion to hold on this forum... :-(
Holding the wheel isn't about how dumb AP is, it's about deflecting liability away from Tesla. It would be the equivalent of having normal cruise control in a car, and then requiring the user to keep their foot on the throttle the whole time or it disengages. Which would make control control absolutely useless. And yes, you can perfectly well crash into something with normal cruise control in a normal car engaged.
AP 1.0 is just providing cruise control+, where it's supposed to steer and brake for you in the dumbest way possible (which is a far cry from what was promised, but fine). You don't need to put your hands on the wheel but you need to pay attention. I've taken away control many times when my hands weren't on the wheel.
Tesla has been continually walking back features since 7.0, not going forwards. Every new feature has been a parlor trick. Like summon.
The scariest aspect of AP1 is that it can't see cars turning in front of you. I have had several cars turn right in front of me and autopilot doesn't sense it, doesn't slow down etc. The other part that is scary is that if a car in front you is at a complete stop it takes a long time to recognize that and slow the car down in time. I drive on a divided highway with a speed limit of 60 mph. This particular highway has stop lights every few miles. If my tesla isn't tracking a car in front of me then I always have to take control of the car because it would not see a stopped car in time to break from 60 mph in time.
AP1 though is still best in class for now until AP2 is fully up and running. We just bought a Pacifica minivan and the one that we test drove had "lane keeping" and it was laughable how bad it was. The thing would drift over to one side and then jerk the car back over to the other side and rinse and repeat. The salesman thought it was the best thing ever and I just sat there thinking that this is what a drunk tesla would be like if cars could drink
Well my earlier point is that with the suite of hardware that is AP1 the car should never have had some of the capabilities that it did in the first place. This system should have never been billed as something that could be used on two lane highway in my opinion. It should have been expressway only, no roads where anything can turn in front of you.
Agreed on Summon. It's a joke and has practical uses for almost nobody and tends not to work for the few people that actually need it. For instance, my house has a carport and to fit both of our cars into it I need to hug the side of the garage. So if I want to pull my daughter out of the car theoretically I could line the car up, remove her, close the door and then use the FOB to summon the car the 5 feet back to the wall of the carport. But since our carport is on a slight decline it works maybe 15% of the time. Situation 2: I'm at my inlaw's house where they have a large motorhome that I have to hug in the driveway. Same issue with getting my daughter out. Same issue with Summon not working but they have a perfectly flat driveway. The problem here is that the car doesn't want to go over the curb. So two of the few situations where one would want to use summon, both of which the feature doesn't perform.
AP1 though is still best in class for now until AP2 is fully up and running. We just bought a Pacifica minivan and the one that we test drove had "lane keeping" and it was laughable how bad it was. The thing would drift over to one side and then jerk the car back over to the other side and rinse and repeat. The salesman thought it was the best thing ever and I just sat there thinking that this is what a drunk tesla would be like if cars could drink
AP1 is amazing and still best in class. It's problems are a result of people thinking it's something it's not, thinking it's more advanced than it is, and Tesla failing to educate owners so they don't make the first two mistakes.
There's no reliable way to do this, more than there is a reliable way to make cruise control only work on the highway.
I don't get this : OP drives at a very unsafe speed of 90 mph way way above the speed limit with AP on. At that speed the AP system is highly stressed in maintaining the lane. That is an abuse of AP. If there had been an accident because lane keeping failed, all the whiners will be up in arms.
So what's wrong is disabling AP and forcing the user not to abuse the system?
I understand that simply canceling AP and not disabling is a choice. But trying to fault them for disabling it because you forced the system to go over 15 or 20 miles over the limit , is silly.
If you want to have fun, then don't force AP to be your co driver
So what's wrong is disabling AP and forcing the user not to abuse the system?
I understand that simply canceling AP and not disabling is a choice. But trying to fault them for disabling it because you forced the system to go over 15 or 20 miles over the limit , is silly.
If you want to have fun, then don't force AP to be your co driver
SomeJoe7777
Marginally-Known Member
You are correct, Texas law 545.157 (Passing Authorized Emergency Vehicle) states that you must move over one lane or slow to 20 MPH below the speed limit when passing an authorized emergency vehicle that is stopped on or beside the roadway.
I would have moved over but there was another vehicle in position to my right such that a move over to that lane would have been too close to him. However, now that I know that Texas has a specific law on the books that covers this situation, I'll be more proactive going forward.
I do not believe AP was reacting to the emergency vehicle in this context. It responded too late, for one thing, and given that AP has difficulty recognizing simpler items like speed limit signs, I can't imagine that Tesla would have attempted to make it recognize emergency vehicles (if the MobileEye chip even has facilities to make that possible, which I doubt).
I believe this is a radar-based speed differential between your car and cars in the adjacent lanes. The radar on the car is not doppler-based so it can't read speeds of objects on a single radar return. But using two time-independent returns can yield speed data, and this fits nicely with Elon's description of using the radar in a temporal fashion to build a 3D object representation like a "poor man's lidar". I believe the intent is to prevent you from zipping along at 70 MPH in a lane adjacent to a large row of cars that is creeping along at 10 MPH. On autosteer, this presents a danger equivalent to using autosteer on a 2-lane road -- the only thing preventing you from colliding with a car in an adjacent lane is flawless camera analysis of the lane stripes, which we all know is not entirely flawless.
mblakele
FSD Beta (99)
Yes, holding the wheel is probably about liability — but I honestly believe it also improves safety by improving engagement. Read enough about airplane accidents and their causes, and I think you'll agree that engagement is important for reducing the risks associated with autonomous systems. Until we can get beyond L2, driver engagement is critical for safety. I'm certain that Tesla's AP team is aware of this.
Because of the importance of engagement, I think your analogy with cruise control is mistaken. Cruise control requires me to remain engaged through steering: its main limitation becomes a safety feature, at least barring "highway hypnosis". Autosteer reduces the functional need for engagement, so we want to stop holding the wheel. But Autosteer isn't L3 or better: it can't handle all situations, nor warn the driver to take over with sufficient lead time. In this situation engagement becomes very important for safety. Engagement reduces mode error, and also reduces the handoff problem.
Keeping a hand on the wheel, even when I don't have to steer, gives me immediate tactile feedback on whatever Autosteer is doing. Even with my eyes on the road, that extra feedback helps improve my reaction time to unexpected events, which improves safety. So I believe that the extra feedback from keeping a hand on the wheel is a net win for driver engagement and safety.
Similar threads
