commaai
Car Not Resuming After A Stop
Describe the bug
While OP is engaged and comes to a stop when a lead car is in front, sometimes it will just sit stationary and not move at all once the lead car moves.
What hardware does this issue affect?
comma two
Provide a route where the issue occurs
69567f65ae707320|2022-05-01--19-28-38--3
openpilot version
0.8.13
Additional info
I understand Toyota's have bad tuning when it comes to Stop and Go longitudinal, but in the route posted there are multiple times where OP just won't start driving again, no matter how far the car is in front.
2022 Camry Hybrid XSE
On the UI screen, the white path planner will also not show once the lead car starts driving.
Timestamps of OP not reengaging:
15:32:34 – 3 15:34:20– 5
It looks like there's a radar track that's reporting an object very close to the car and the low speed override keeps you stopped for it. @HaraldSchafer Is this something radard should ignore, or is it working as expected and the issue is the radar?
Interesting. Could this be some type of object on or near my radar that I need to remove?
On Mon, May 2, 2022 at 4:41 PM Shane Smiskol @.***> wrote:
It looks like there's a radar track that's reporting an object very close to the car and the low speed override keeps you stopped for it.
[image: Screenshot from 2022-05-02 13-40-07] https://user-images.githubusercontent.com/25857203/166324001-dca34b7b-0ab1-4dd8-b64a-1c9609492052.png
— Reply to this email directly, view it on GitHub https://github.com/commaai/openpilot/issues/24395#issuecomment-1115345903, or unsubscribe https://github.com/notifications/unsubscribe-auth/AH6CEUGF4ROQYVWTYJJFCLLVIA4W5ANCNFSM5U44FC5A . You are receiving this because you authored the thread.Message ID: @.***>
--
James LaRosa | CEO Phone: (304) 460-9719 Location: West Virginia
Hmmm, this is the intended behavior the way radard works now. I haven't seen this before where the radar persistently predicts an object super closeby . Can you try visually inspecting your radar sensor and cleaning it? If you still have the problem after I will consider changing radard to be robust to this.
@HaraldSchafer sorry for the late response. I am still getting the same thing happening. Over the last week it got better, but had a 20 minute traffic jam today and it consistently kept happening.
https://share.icloud.com/photos/03eKNdWzgGh_nEu96nKH-g9tw
this was from today. I am running DP so I can get the do not cancel on accel since the stop and go is really bad. Are you able to use any of this data or should I reinstall stock and send over a cabana link after I get a drive in?
The problem is still persisting. Another user in the Discord described an issue that this also contains, starting from a stop then randomly jamming the breaks and stops moving. Here is the Discord message: https://discord.com/channels/469524606043160576/524327905937850394/987482601658937385
I was running the e2e branch today, but the issue was still present. The following routes are when I experienced the issue:
76fcc86c5aa85c45|2022-06-21--16-58-19--4
76fcc86c5aa85c45|2022-06-21--16-58-19--5
76fcc86c5aa85c45|2022-06-21--16-58-19--6
76fcc86c5aa85c45|2022-06-21--16-58-19--12
I also inspected the radar yesterday as well as went to a car wash and sprayed off anything that could've been on the radar. If you guys need any more data from me please let me know and I will be happy to help get this issue fixed!
I’ll get some data today on Master and put the routes in here. It happens more on E2E vs Master (assume because the king is very experimental on that branch) but still happens on Master
@HaraldSchafer I think the Toyota radars stop reporting points at some distance, I'll see if it's around 0.5
@sshane posting this here as well in case anyone else wants to look in on it. On your remove-toyota-deadzone branch, had another instance of it not resuming after a stop. Dongle ID: 76fcc86c5aa85c45 | route: 76fcc86c5aa85c45|2022-06-25--14-03-07--8
also, right after that specific scene in the route, my C3 froze and then force restarted itself. Not sure if it’s related to the branch I was testing out or another issue.
Tested out remove-toyota-deadzone branch again today. Another instance of OP being at a stop, the lead car resuming, and OP sits at a standstill
Dongle ID: 76fcc86c5aa85c45 | route: 76fcc86c5aa85c45|2022-06-26--10-51-47--4
Master-Ci, same issue, not resuming after a stop
Dongle ID: 76fcc86c5aa85c45 | route: 76fcc86c5aa85c45|2022-07-04--15-07-34--4
Master-CI, not resuming after a stop:
Dongle ID: 76fcc86c5aa85c45 | route: 76fcc86c5aa85c45|2022-07-21--21-12-16--63
Master-CI, few instances of it not stopping behind a car:
Dongle ID: 76fcc86c5aa85c45 | route: 76fcc86c5aa85c45|2022-07-27--15-32-54--6
Dongle ID: 76fcc86c5aa85c45 | route: 76fcc86c5aa85c45|2022-07-27--15-32-54--7 (my car started to go after a stop, but then hit the breaks randomly and wouldn't continue)
Dongle ID: 76fcc86c5aa85c45 | route: 76fcc86c5aa85c45|2022-07-27--15-32-54--7
Dongle ID: 76fcc86c5aa85c45 | route: 76fcc86c5aa85c45|2022-07-27--15-32-54--9
Master-CI (0.8.16), heavy stop and go traffic, and problem of not resuming after a stop still persisting:
Dongle ID: 76fcc86c5aa85c45 | route: 76fcc86c5aa85c45|2022-07-30--14-39-44--14
Dongle ID: 76fcc86c5aa85c45 | route: 76fcc86c5aa85c45|2022-07-30--14-39-44--15
Dongle ID: 76fcc86c5aa85c45 | route: 76fcc86c5aa85c45|2022-07-30--14-39-44--17
Dongle ID: 76fcc86c5aa85c45 | route: 76fcc86c5aa85c45|2022-07-30--14-39-44--18
@JamesL6 can you try openpilot disabled (dashcam mode) and see if you can reproduce this in stop and go with the stock system?
@sshane For sure, will try in the next few days. I have used the stock system before and never got this issue, but will see if I can reproduce it.
@sshane Using Stock Toyota:
Dongle ID: 76fcc86c5aa85c45 | route: 2022-08-09--16-54-45--0 Dongle ID: 76fcc86c5aa85c45 | route: 2022-08-09--17-37-28--0
@sshane A lot of successful stop and go's in this drive, using stock Toyota.
Dongle ID: 76fcc86c5aa85c45 | route: 2022-08-12--13-28-11--16
Testing @sshane toyota-maybe-good-long
I had some more successful stop-and-go's with this branch than compared to Master or E2E Long. When it would successfully accelerate from a stop, most of the time it was too slow to accelerate, the lead car would then leave a huge gap between us, then OP would accelerate too much, just to hit the brakes again in traffic
It also did not work 100% of the time, and the issue of not resuming after a stop was still happening.
Dongle ID: 76fcc86c5aa85c45 | route: 2022-08-20--18-40-25--0
Dongle ID: 76fcc86c5aa85c45 | route: 2022-08-20--22-17-34--4
Dongle ID: 76fcc86c5aa85c45 | route: 2022-08-20--19-15-10--17
Used the new Master 0.8.17 Update today on several drives. I did not encounter 1 instance of Openpilot not resuming after a stop and go, it worked as it should.
Dongle ID: 76fcc86c5aa85c45 | route: 2022-08-25--19-41-51--0
Dongle ID: 76fcc86c5aa85c45 | route: 2022-08-25--16-23-47--0
Master 0.8.17 is now not resuming after a Stop in most scenarios. I will upload routes later.
Master e2e long toggle enables (i know it is highly experimental, but e2e long has the highest chance of actually resuming from a stop than any other branch, so throwing this in here)
Dongle ID: 76fcc86c5aa85c45 | route: 2022-09-09--20-50-31--3 (rolls back on a hill and doesn't resume from a stop) Dongle ID: 76fcc86c5aa85c45 | route: 2022-09-09--20-50-31--11 (was showing the lead car triangle on this road, like it was seeing something)
This is happening in VW Atlas 2018 as well.
VW uses stock ACC at this time (though OP long support is near!) so this potential Toyota/OP long issue is unrelated.
There was indeed a regression in cancel/resume reliability in openpilot 0.8.16, but it is resolved in #25692. The fix will be included in openpilot 0.8.17. If you'd like to test the fix right away, you can install master-ci.
I will check this out next week, in the mean time can you make sure a lot of examples are fully uploaded with rlogs so I can quickly check it out?
I believe the majority, if not all the routes, I uploaded all the files/rlogs 👍
Will go for a spin today too and post a route here