My team currently has one Limelight 2+. From my understanding, it is relative simple to get the 3 to recognize April tags, and next years game will have only April Tags and no reflective tape (FRC Blog Post from a while ago). How much extra work is it to get the 2+ recognizing April tags and is it enough to justify purchasing the 3 to avoid it? Any advice would be appreciated.
The work is the same to get a 2 or a 3 to recognize april tags. Upgrade the firmware.
The difference is in performance and the 3 will give more frequent april tag positions. If you are going to use the tags for latency adjusted localization in addition to odometry (see pose estimation) a 2 is probably fast enough, if you are trying to allign on goal using an april tag a 3 will serve you much batter.
My teams experience with vision was a massive journey. With the 2 we had apriltag recognition out to 10 ish feet consistently. however this recognition was iffy at best and when used to find the bots field position in limelight it barely worked, we had problems with the limelight thinking our robot was in the floor, and could only give us a general idea where we were on the field. With 2 limelights crossing each others field of vision we could fix this issue, but i dont know if your team has 2. I dont remember our precise results on getting apriltags with a limelight 3 but I do remember it was much more accurate at range. A limelight 3 also really opens up options like detecting game pieces and using that information in a match.
Cool. I’ll definitely try to push for our team buying one then. Thanks for both of the responses!
Limelight 2 can easily locate the purple object or the yellow object with HSV thresholding.
The 2+ supports the Google Coral as well. I can’t remember if the 2 supports it but it might not.
They can yes, however we got a massive increase in performance when we upgraded. Much more reliable tracking of apriltags, and larger detection range of game pieces. You can do all these things with a 2 but the 3 is both easier to use and is better at everything wanted than the 2.
The LL3 will find tags about 2x as fast as a LL2. A custom solution is more work but is 2x faster than a LL3. I can’t comment on the relative camera quality of the two on a robot, but both do work about the same on the bench (other than the 2x performance difference).
More information here: Coprocessor - Google Drive
This topic was automatically closed 365 days after the last reply. New replies are no longer allowed.