TMC is an independent, primarily volunteer organization that relies on ad revenue to cover its operating costs. Please consider whitelisting TMC on your ad blocker and becoming a Supporting Member. For more info: Support TMC
Start a Discussionhttps://teslamotorsclub.com/tmc/tags/

AP1.0 Reducing Speed for, Red Light, "T" Intersection - VIDEO

Discussion in 'Model S' started by JohnnyG, Apr 18, 2017.

Tags:
  1. JohnnyG

    JohnnyG Weee!

    Joined:
    Jun 24, 2016
    Messages:
    445
    Location:
    Columbus, OH
    2015 Tesla Model S Autopilot HW1 17.11.10

    Interesting reproducible behavior while approaching a red light 3-way intersection. The Autopilot is reducing the speed as it approaches the red light at the 3-way intersection. Nothing appears on the IC denoting the traffic light nor the T-intersection.

    Notes: I correct the steering in the apex of the sharp curve because AP goes wide, however the TACC never disengages. I immediately re-engage the AS upon exiting the curve. It is after this that the system begins slowing the car. As I get close to the blind intersection, I retake control and stop the car, because I don't know if the car would have completely on it's own.

     
  2. MarcusMaximus

    MarcusMaximus Member

    Joined:
    Jan 2, 2017
    Messages:
    592
    Location:
    San Jose
    Offhand, that looks more like the typical reaction of AS/TACC when approaching a tight turn, rather than anything to do with the light.
     
  3. Max*

    Max* Banned

    Joined:
    Apr 8, 2015
    Messages:
    6,425
    Location:
    NoVa
    The only way it proves that it's a red light effect, is if you have a video of it NOT slowing down for a green light, and both experiments are repeatable.. Otherwise I agree with @MarcusMaximus

    Also, it didn't slow down enough to stop at the light, so it's doubtful it's the light causing the slow down.
     
  4. JohnnyG

    JohnnyG Weee!

    Joined:
    Jun 24, 2016
    Messages:
    445
    Location:
    Columbus, OH
    #4 JohnnyG, Apr 18, 2017
    Last edited: Apr 18, 2017
    It slowed *after* the curve... and continued to decelerate while the road was straight. The road ahead, at this point, did not have a tight turn.
     
  5. JohnnyG

    JohnnyG Weee!

    Joined:
    Jun 24, 2016
    Messages:
    445
    Location:
    Columbus, OH
    #5 JohnnyG, Apr 18, 2017
    Last edited: Apr 18, 2017
    I (partially) agree about the green light, unfortunately this light is not a timed light; it's a triggered light. It would be extremely difficult to catch it just right, and not have a car there; but I will try later this week.
    I think it has more to do with the T-intersection (road ending). It was already past the curve, so it had nothing to do with the tight curve.
     
  6. MarcusMaximus

    MarcusMaximus Member

    Joined:
    Jan 2, 2017
    Messages:
    592
    Location:
    San Jose
    Either I'm misunderstanding or I didn't word my response clearly enough. I mean the turn implicit by the T-junction. Either way you would have to turn tightly at that intersection. I suspect the car was just reacting to that and slowing down so you didn't plow into the sign on the other side.
     
  7. JohnnyG

    JohnnyG Weee!

    Joined:
    Jun 24, 2016
    Messages:
    445
    Location:
    Columbus, OH
    #7 JohnnyG, Apr 18, 2017
    Last edited: Apr 18, 2017
    MORE INFO: This is a common route home for me. Prior to v8.1 it *always* continued to maintain 45MPH after the curve. This is new behavior. I've been able to reproduce it multiple times.
     
  8. chillaban

    chillaban Active Member

    Joined:
    May 5, 2016
    Messages:
    2,059
    Location:
    Bay Area
    I suspect this is radar whitelisting being concerned about the metal traffic sign that you would otherwise be running into…
     
    • Like x 1
  9. JohnnyG

    JohnnyG Weee!

    Joined:
    Jun 24, 2016
    Messages:
    445
    Location:
    Columbus, OH
    I tend to agree more with this theory. This is still new behavior though, and not related to the already existing behavior of attempting to slow for curves (which always seems to slow after you're already in the curve, instead of way before).
     
  10. JohnnyG

    JohnnyG Weee!

    Joined:
    Jun 24, 2016
    Messages:
    445
    Location:
    Columbus, OH
    I hadn't thought about radar whitelisting. That would also explain nothing appearing on the IC.
    Still interesting (odd?) that it slows so early and so gradually. Not (previously) normal behavior for AP with obstacles. Definitely a nice change though.
     
  11. chillaban

    chillaban Active Member

    Joined:
    May 5, 2016
    Messages:
    2,059
    Location:
    Bay Area
    It seems like for obstacles confirmed by the camera, the car takes much more decisive and abrupt action. For radar, AP likes to just coast or slow down a bit, possibly buying the camera more time to react, or hoping it'd bring your attention to the road again.

    Elon said in the 8.0 announcement they are making the radar a primary sensor just like the camera. Well… I dunno, I'd say they made it like a tier 1.5 sensor. Before, if radar sees something but the camera doesn't, AP1 will flat out ignore the radar and trust camera. Now, radar can play a role in causing braking even if the camera doesn't see anything. I would guess the degree of braking depends on the whitelist confidence.
     
  12. ZERO260

    ZERO260 Member

    Joined:
    May 22, 2016
    Messages:
    90
    Location:
    Waldorf MD
    I think it relates to GPS. I think the map indicates a t-intersection and knows that a reduction in speed is required to make R or L turn. This occurs when I a approach a certain intersection on my morning route AP auto slows down whether the light is green or read.
     
  13. JohnnyG

    JohnnyG Weee!

    Joined:
    Jun 24, 2016
    Messages:
    445
    Location:
    Columbus, OH
    Hmm, interesting too... Do you notice this behavior at all T-intersections or just this one?
     

Share This Page