Perhaps I misunderstand, but does AEB not kick in if I fail to react to the FCW alert?
-
Want to remove ads? Register an account and login to see fewer ads, and become a Supporting Member to remove almost all ads.
2017.42 a88c8d5
- Thread starter Peteybabes
- Start date
They are separate and either one, or both, can be turned off. FCW can be set to alert "early", "normal", or "late". if you have FCW set to "late" and you don't react, and conditions don't change, AEB would likely trigger very quickly.
Think of it this way:
FCW: Alerts if it thinks you might hit something and you could do something to avoid the collision.
AEB: Is triggered once it knows that you are going to hit something. (Where nothing you could do would avoid the collision.)
I drove 200 miles yesterday (AP2) both I 5 south of Stockton and Pacheco Pass. It struggled to stay in the lane on sharper freeway curves at 75mph but no diving for the off ramp. Lane change very smooth but would not cut in close to the front of a Semi. It delays the lane change until more distance in front of the Semi than if I were in control. Still unable to handle twisty 2 lane roads or up and down roads where the lane lines are invisible beyond a hill crest.
Okay, that’s how I understood it. AEB isn’t something you’re going to want to test, but you can test FCW and be reasonably sure AEB will intervene and not end up testing your local collision repair shop.
No, AEB will not prevent a collision. It is only designed to reduce the severity of an unavoidable collision.
Sorry, my comment was missing a semicolon. My statement was in response to question about testing AEB. You can test FCW and be reasonably sure that AEB would intervene if you ignore the alert. By the time you find out if AEB is going to kick in, you’re going to collide.
So elaborate on those videos you see on YouTube of our cars doing a swerve around imminent crashes? Because sometimes I’m under the impression the car will swerve to prevent a collision in some cases. Otherwise it may just be the car warning the driver and them swerving manually.
Disclaimer: I may be horribly wrong
Currently the only feature resembling that is side collision assist, which is a narrow domain. It’s strictly triggered by the side parking sensor detecting something a few inches away and that will push your steering in the opposite direction while beeping. There is no other capability in Autopilot or Autosteer to “swerve”. It was on the list of dozens of 8.0 features that were promised in the “Seeing the world in Radar” blog post, but quite frankly, less than half of those items seem like they were implemented. We are still arguing over whether there is any evidence of radar whitelisting happening in the code, the other items seem like they’re more a representation of their medium term roadmap rather than present capabilities.
That's a good question and I don't know the answer. It's my assumption that while driving with autopilot active, the vehicle will slow down to avoid an obstacle and potentially move over within the lane to avoid something coming from the side. It my belief (and my hope) that the vehicle would never leave the lane without manual intervention. I'd rather push somebody back into their lane if they decide to sideswipe me than have the car swerve into oncoming traffic to avoid them. I think that autopilot will stop to avoid a collision if it properly detects an obstacle and has sufficient time to react.
While driving manually, I believe all it will do is warn if it detects a potential collision and brake if it determines that a collision is unavoidable. It's my understanding that if you ignore the FCW beep, that the car will happily crash right into whatever obstacle it detects, only braking at the last moment to reduce the impact speed.
It’s also worth noting that when a collision warning alarm is sounding, the Autopilot disengagement tone is not played when you take over steering. So in a video it’s extremely hard to tell whether it’s the driver responding to a collision alarm or Autopilot doing it on its own. I do have to say, I doubt anyone is insane enough to sit idly by and rely on AutoSteer when there is an imminent collision.
Last night I received the latest firmware being discussed here. I have EAP. Today I tried AP and TACC in the following scenarios:
280 freeway from San Mateo to Page Mill Road (California, San Francisco peninsula area): Within a minute after activating AP on the freeway I realized that AutoSteer was noticeably more stable and was maintaining lane position better whether the road was straight or curved. This was at 0800 and sometimes straight into the sun which was fairly low, a challenging situation for the cameras.
Hwy 35 (Skyline Blvd) from Page Mill Road to Hwy 92: This is a narrow curvy two-lane road, I used TACC for most of it (not AP), set to 45 or 50mph, and was impressed by how the car automatically slowed down for tighter curves and then accelerated after the curve back up to the speed setting. It was really remarkable how good it was! It wasn’t perfect, but it was very good. I’ve driven this road over a hundred times (never used TACC on it before) and in most cases the car took the curves at speeds that I was comfortable with. Sure I would have taken some of the curves somewhat faster, but in general the car got the speed right in a curve.
But here is the really amazing part: multiple times, the car started to slow for a tight curve BEFORE the lane lines on the drivers display showed the curve. The lane lines shown on the display would be straight lines and the curve about a hundred feet away or more and at that distance the car would begin to slow down to a speed that I felt was very appropriate for the curve.
I had not set a destination in the navigation so the car was not following a navigation route. Again, this was with only TACC on, not AutoSteer.
It was as if the car knew the turn was coming up and how severe the curve was without the drivers display indicating that the car was visualizing the curve through the forward cameras.
Is this evidence that Tesla is now using onboard high-precision mapping? Or is TACC using Google Maps to anticipate curves based on the known position of the car on the map? Or both? I don’t know.
Maybe others have already seen this TACC behavior before the latest firmware release (2017.42). I have seen hints of it on 2017.40 but nothing like what I experienced today.
280 freeway from San Mateo to Page Mill Road (California, San Francisco peninsula area): Within a minute after activating AP on the freeway I realized that AutoSteer was noticeably more stable and was maintaining lane position better whether the road was straight or curved. This was at 0800 and sometimes straight into the sun which was fairly low, a challenging situation for the cameras.
Hwy 35 (Skyline Blvd) from Page Mill Road to Hwy 92: This is a narrow curvy two-lane road, I used TACC for most of it (not AP), set to 45 or 50mph, and was impressed by how the car automatically slowed down for tighter curves and then accelerated after the curve back up to the speed setting. It was really remarkable how good it was! It wasn’t perfect, but it was very good. I’ve driven this road over a hundred times (never used TACC on it before) and in most cases the car took the curves at speeds that I was comfortable with. Sure I would have taken some of the curves somewhat faster, but in general the car got the speed right in a curve.
But here is the really amazing part: multiple times, the car started to slow for a tight curve BEFORE the lane lines on the drivers display showed the curve. The lane lines shown on the display would be straight lines and the curve about a hundred feet away or more and at that distance the car would begin to slow down to a speed that I felt was very appropriate for the curve.
I had not set a destination in the navigation so the car was not following a navigation route. Again, this was with only TACC on, not AutoSteer.
It was as if the car knew the turn was coming up and how severe the curve was without the drivers display indicating that the car was visualizing the curve through the forward cameras.
Is this evidence that Tesla is now using onboard high-precision mapping? Or is TACC using Google Maps to anticipate curves based on the known position of the car on the map? Or both? I don’t know.
Maybe others have already seen this TACC behavior before the latest firmware release (2017.42). I have seen hints of it on 2017.40 but nothing like what I experienced today.
Updated on Wed this week and drove the car today for ~100 miles and hands down this is the best I've seen AP2 perform since we bought the car. I'd say this was equal (maybe even better) than the AP1 XP90D loaner we had for 1 months while our car was at the SC getting some fixes. The bad ping ponging I was experiencing in the past several SW updates is pretty much gone and the lane keeping and steering overall was very tight. Lane changing was good (maybe not the best I've experienced in one of the previous builds it was slightly smoother), but it was completely acceptable. This was 98% highway. I also tried it on local roads a few times and that still wasn't quite there, but I also didn't expect it to blow me away. Today's drive showed me promise that over the next few months (hopefully sooner) we will start seeing AP2 surpass AP1.
What still surprises me is the inconsistency of the results between cars that are both AP2 on the same build. One side note is that my windshield was just replaced this week and at first I thought they didn't recalibrate it and it was hugging the right side of the lane (on Wed) and then after driving the car Thurs/Fri, apparently, this morning when I tried AP, it worked just fine. So it must have calibrated itself. I'm curious if others have any background on how the calibration works or if recalibrating AP actually clears some previous history and lets you start fresh in your AP driving experience?
What still surprises me is the inconsistency of the results between cars that are both AP2 on the same build. One side note is that my windshield was just replaced this week and at first I thought they didn't recalibrate it and it was hugging the right side of the lane (on Wed) and then after driving the car Thurs/Fri, apparently, this morning when I tried AP, it worked just fine. So it must have calibrated itself. I'm curious if others have any background on how the calibration works or if recalibrating AP actually clears some previous history and lets you start fresh in your AP driving experience?
Interesting. Another option is that the display on the dash showing the upcoming curve is behind (delayed) what the cameras are actually seeing? The camera "sees" a lot quicker/further than you think and starts to take action before you see it on the dash. Could also be a combination of this and google maps?
Either way this is great. Starting to see real progress on AP2 and hopefully will continue to get better and better.
Either way this is great. Starting to see real progress on AP2 and hopefully will continue to get better and better.
Interesting, have to check this out. I’m sure you would have mentioned if you had ‘blue car syndrome’
and were locked onto the car in front of you. Since not, it has to be advance camera use or mapping. I always wondering why maps weren’t used to tell when to decelerate in advance.
Sorry, I should have been clear that the behavior I described — TACC appearing to know a tight curve was coming up before the drivers display showed curving lane lines — happened when there was no car in front of me.I’m sure you would have mentioned if you had ‘blue car syndrome’
I cannot rule that out. But if that is the case, why would Tesla choose not to show that data to the driver? If the car is taking action based on knowing a tight curve is coming, why not show the curve on the display? Doing that would reassure the driver that the car is slowing for a good reason. On the road I was on, if it was night and the driver was not familiar with the road, the driver might wonder why the car was slowing and might override it because they couldn’t see the curve yet (there are no streetlights by the curves I was describing).
There's Tesla's Autopilot described. Just look how they describe it still today... not much resemblance with the software shipping in the car... Autopilot
Tozla
Member
I've noticed that this update has a more sensitive forward collision warning. I'm getting the alert more frequently and my driving habits haven't changed as far as I'm aware.
Wow, I've never actually read that before. Probably a good thing, because I might be unhappy with the current state of autopilot if I had...
Similar threads
- gogman69-rip@y
- Model X: Ordering, Production, Delivery
