Conversation
| Logger.recordOutput( | ||
| cameraHelper.getLogPath() + "/RobotPoses", | ||
| cameraHelper.getRobotPoses().toArray(new Pose3d[0])); | ||
| Logger.recordOutput( |
There was a problem hiding this comment.
We shouldn't log these by default. As the robot season progresses, we frequently end up with too much data in NetworkTables which impacts robot performance, so we need to be careful about logging large data structures.
There was a problem hiding this comment.
I understand that we have to be frugal with our NetworkTables usage, but this data (what we're rejecting and why) is the starting point to address issues like deadspots or ghosting. Is there a better place to record debug data that doesn't bottleneck the robot? Or do we just hope that we catch any issues on the practice field, and not record it during the real competition matches (which arguably are an environment that isn't quite reproducible elsewhere, at least not in quantity)?
There was a problem hiding this comment.
Oh or are you saying don't record allPoseObservations and acceptedPoseObservations because that's duplicating data, and I could just do acceptedPoseObservations and rejectedPoseObservations so the same observation is never recorded twice?
There was a problem hiding this comment.
in addition to not double-logging everything, we could add a paramter to control the log level of these, so that with a quick deploy we can turn them on or off (without having to change SCL)
pwbriggs
force-pushed
the
pwbriggs/more-vision-logging
branch
from
6d14d96 to
7082dd4
Compare
pwbriggs
force-pushed
the
pwbriggs/more-vision-logging
branch
from
7082dd4 to
c9ffe82
Compare
| Logger.recordOutput( | ||
| cameraHelper.getLogPath() + "/RobotPoses", | ||
| cameraHelper.getRobotPoses().toArray(new Pose3d[0])); | ||
| Logger.recordOutput( |
There was a problem hiding this comment.
in addition to not double-logging everything, we could add a paramter to control the log level of these, so that with a quick deploy we can turn them on or off (without having to change SCL)
pwbriggs
force-pushed
the
pwbriggs/more-vision-logging
branch
from
c9ffe82 to
94ab67f
Compare
|
sorry I rebased the wrong branch. I'll clean this up later |
pwbriggs
force-pushed
the
pwbriggs/more-vision-logging
branch
from
94ab67f to
7277998
Compare
3 participants
Why are we doing this?
This makes SCL record the data I wish I had when I was debugging AprilTags over the summer.
What's changing?
Record the poses that we reject, as well as more info about each tag observation (ambiguity, multi-tag count)
Questions/notes for reviewers
How this was tested