Limelight Accuracy Issues

Technical Programming
1

Looking for help/suggestions for a problem where we are stumped. We are using a limelight 3 and have been looking at the accuracy of the data from the limelight based on the apriltags. We positioned the robot at two different positions, both straight on. One position is at about 100 cm and the other is about 290 cm. See the picture below for the details.

Our limelight is mounted at the back of the robot, so we have the yaw set to 180 degrees. The camera is tilted so the pitch is 40 degrees and the roll is 0 degrees. The limelight position basically is 16 cm up, -35.5 cm back and 0 cm across.

EDIT: Yes the setting in the Web UI is 0.16 and -0.355.

Our driver station is set to blue and we are sending the correct (0 degrees) offset to the mega tag 2 via the SetRobotOrientation() API. We are providing the yaw value and zero for all of the other values. The robot is completely stationary.

From the picture below you can see that the original mega tag is ok at 300 cm, but is off more than I would expect (9.7 cm) at 100cm. In both cases the mega tag 2 is off a significant amount.

Any ideas? I have read somewhere that the straight on problem is more difficult to solve and we will try at an angle tomorrow, but it seems while sitting completely still that the values would be more accurate than what we see.

I noted that the limits for camera roll are +/- 44 degrees, are there any limits for pitch?

We have tried changing the limelight for a different with the same results.
We have tried changing the frame of reference so that the camera looks like it is one the front of the robot.

Note, we are reading values via the LimeLight web interface and not looking at the robot pose after vision samples are fed back into the swerve drive odometry.

Problem Diagram

Thanks
Butch Griffin
Error Code Xero, 1425

2

What Gyro are you using, and are you allowing to time to calibrate before moving the robot?

MegaTag2 is fairly reliant on Gyro integration, so if there’s any issues or unexpected things happening with your Gyro it could be causing issues.

We actually reverted back to MegaTag1 part way through the season because of issues we kept having with MegaTag2.

3

We are using the Pigeon 2, but given our orientation we just forced a 0 degree drive base heading as one of our experiments with no impact on the results.

4

Resolved. This was due to the configured apriltag size slider not matching the 2024 tags.

I think we will warn users if the slider does not match the configured map tags. Thanks for walking me through your setup and for the suggestion @sjcbulldog

1 Like
5

Thanks for all your timely assistance.

1 Like
6

Note that you should calibrate the LL. If you don’t, your pose measurements will be off by 5 to 25% due to lens and sensor variations.

1 Like
7

2024.10.1 includes a new warning to help guard against this.
Thanks @sjcbulldog !

2 Likes
8

Excellent. Will save the next team hours :). Not complaining though, we greatly appreciate what you do. In 2019 (Hatches + Cargo) we did our own vision (Raspberry PI + USB Cameras) and it was a very difficult thing to get done. We fortunately had one of these students that was tremendous and I hired him as an intern the minute he graduated high school (and before college). Without him it would have been difficult. The limelight allows me to teach students how the complete system works without having to deal with the hundreds of small issues that would detract from that.

4 Likes