Ford: use more reliable gear message #32886
Conversation
selfdrive/car/ford/carstate.py
Outdated
| # TODO | ||
| # - [x] NOPE: PowertrainData_10->GearLvr_D_ActlDrv (goes into L instead of D for the rest of the drive randomly, commonly when current message goes into unknown gear) | ||
| # - [ ] PowertrainData_10->TrnRng_D_Rq | ||
| # - [x] NOPE: TransGearData->GearLvrPos_D_Actl (goes into N instead of P, this is the transmission gear) |
There was a problem hiding this comment.
The signal previously used before adding the Lincoln Aviator in #27095 also goes to N around the time that these three candidates do, in cases where the current signal does not. It seems we are currently missing some cases where the user or car puts the gear into N which the camera listens to and disengages, so this seems like a good all around change.
The new signals (and the one before the current one) also switch to N when switching gears which makes sense if it's the actual gear of the transmission/powertrain system. It looks like the current signal is a gear request from the gear shift module (Gsm), which might not be the true gear.
(third plot from the top: orange is the previous signal, blue is the current signal)
|
Some mismatches greater than 0.1s, in all cases the new signal ( ViewHere the PCM cruise state signal fell when the PCM gear went to N, but this wasn't captured in our current gear request signal. For segments with mismatches greater than 0.02s, it appears the common theme is that this new signal goes to N on changing edge of gears or just switches earlier: View
|
|
The current signal also does not capture quickly changing into N and back to D: View
Process replay diff is expected, we get this View
|
* todo * nope * not this one * yep * passes * clean up * not used! * oop * update refs old-commit-hash: 7de2aac
Confirmed this new signal does not have the problem in the Aviator described in #27095