Hello,
For the next season, we plan to use limelight 3G for vision processing because we want to make it easier as a rookie team. For an average case, what is the optimal number of webcams/limelights we should use?
Thanks
Unless you have something specific in mind, start with two - one to mount on the robot, another for experimentation offboard and to be your spare in competition.
The real number to mount on the robot depends on what you want to do, and how gamepeices and tags are arranged on the field relative to the positions of importance.
Having two in hand before the season also lets you mount two if you want, and order a 3rd spare if needed.
2 Likes
We used 4 Limelights this past year for apriltag tracking and were discussing adding a 5th for game piece tracking.
That being said, as a rookie team, I’m not sure you should try to tackle Apriltag position tracking your first year. Maybe use one for targeting (and maybe an extra as a spare), but going full position tracking right off the bat is probably a bit ambitious for a new team.
3 Likes
It depends on what you want to do, and how far you’re willing to go.
For simple vision targeting, a single one is likely enough.
Maybe you want to simultaneously do object detection, so you’d want another one there.
If you want it to aid in pose estimation, then depending on how much you care about knowing your pose exactly at all times, you can run more or less cameras. If you only care in auto, then a single might be good enough. 3005 as far as I know only had a single camera.
But having more for AprilTag pose estimation can give you better results. 2 is as low as I’d go for that purpose personally, but more can be better, but the returns are less for each additional one you add.
1 Like
+1. Full field pose estimation is likely a “prove it works over the summer” project still - use caution assuming it can be done successfully in season.
What I think would be tractable within build season - Visual servoing - basically, just turn left or right till the apriltag is centered in the camera image. If the GDC holds course, there will be at least one game component where simply “turn toward the apriltag” will get you pointed correctly to pickup, score, etc.
3 Likes
This is always going to depend on what you want to do with it, what you can do with it, and specifics about the game and your robot.Are you planning to do pose estimation? Do you have a turret? Is there an April tag on every location you want to score at or do you need multiple limelights pointing everywhere.
Our experience has been one limelight for pose estimation was plenty for our use case this year. Especially having since tried megatag2 you don’t really need more then 1. The only thing is if you want to focus on specific April tags and you can score on multiple sides of the robot. This year for example some robots have a shooter on one side and an amp mechanism on the other side. If you want to see the April tags on both the speaker and amp when approaching them then you need two cameras.
Don’t over think things though. It’s easy to say more cameras will give us better results but your also just creating more work that may be better spent elsewhere.
Also think if you want to do game piece sensing. Then you would want a limelight 3. We had one 3g and one 3 on our robot.
Also the new radios have 4 ethernet ports. One goes to the Rio. One is for diverstation. And two are for other devices like limelights. If you have more then two limelights/ethernet devices then your going to need a network switch. I’d try to stick to two limelights if possible.