If so, what were your experiences with it? Our time at a practice field is pretty few and far between, would it be a worthwhile investment given that?
We used it for cube tracking last year and it worked pretty well, but it was never really needed once our intake was good enough.
Mostly it was used for the drive cams.
This depends entirely on the game and isn’t something I can really answer now.
If it was like this year’s game where you could use a single game piece and track it in a line to test it you may not need a practice field or much space at all.
If it is like 2017 you still may not need a practice field, but may need a bit more space to be able to replicate a true vision target.
(this was our setup in 2017, before we had a practice field at the school)
I have been a Beta Tester for Lime Light the past two years. The best answer for both veteran and experienced teams looking to put vision on their robot is YES Lime Light is worthwhile investment. Last years game was not a very good game to judge vision requirements. Seeing that the materials list suggests up to 14 feet of 2" reflective tape for just the half field we can safely assume there will be a significant vision target this year. The development time to get a reliable vision tracking system installed and tracking targets with Lime light will be under an hour. So with two bolts, One Injected POE cable, and my Lime Light library. Your robot will be scoring unobtainium rapidly and reliably in no time.
Same on the beta testing. Same assessment of the LL for teams as an investment. It’s worth it.
I’ll caveat this and say that we haven’t used the LL at an event yet but we’re doing other things and just because we are or aren’t doing something isn’t a reason for anyone else to do anything. We’re a bit odd. The Limelight is awesome and a lot more teams should be looking at it or using it.
My biggest issue with the LL, isn’t with the LL, it’s with FRC and our continued insistence that we all use NetworkTables instead of ZeroMQ or OSC (or even ROS messages but that is another story). NetworkTables wasn’t great to begin with and I’m just not a fan of how it is architected* - even after the big changes to it. All that said, it works for most teams just fine.
*I’m nearly at a point with NT that if someone came along with a ZeroMQ or OSC library project for FRC then I’d probably throw some time and money at it to help them because I feel like it’s worth moving away from NT.
As previously mentioned, the LL wasn’t of much use for vision last year due to the lack of vision targets. We tried using it for tracking cubes and performing automatic pickup but due to the varying orientations of the cube, we could never really pull it off, unlike a vision target such as on the 2016 Stronghold goals.
The LL is extremely useful for a plug-in camera because it was extremely rare that we weren’t able to access the live camera feed from it. Compared to previous driver cameras, the LL was very reliable; it even ensured the reliability of a second camera that was plugged into it. We found this particular feature to be very useful.
One issue I found while trying to do vision with the LL is its frequent troubles with network tables. Often, the LL would not even publish values that were necessary for vision to the network tables visible on the driver station, and we couldn’t even turn the LEDs off at one point (when we decided to use it as a driver camera). Updating the firmware didn’t do much to fix this. At some point in between, this issue was fixed, but then the patch somehow was overwritten and we were once more unable to see any published values.
Note to my final comment- I haven’t tried out the Limelight in a month or two, so it’s possible that LL has released a new firmware that fixes this issue.
I’ve heard this sentiment before, and would you mind explaining why NT isn’t the greatest? I like zeromq personally but always assumed NT was the default go-to for FRC. Thanks in advance.
We bought ours for the vision tracking but ended up only using it for the live video feed. That said, with two cameras (the Limelite plus a USB camera connected to it), we got far better framerates at higher resolutions than any other solutions I’ve come across.
Im going to say things now and get in trouble for being opinionated.
- It seems highly susceptible to interruptions and intermittent disconnection issues (much more so ZeroMQ).
- It started life as a hack that has been refined instead of being a purpose built solution from the onset.
- It teaches students some bad habits for network communication.
- I’ve had a lot of bad experiences with it over the years (yes, we’re still using it for dashboard stuff, even today).
- I’d much rather see it replaced with an open and robust solution that has better support for the kind of customizations that I think people want instead of the continued hackery we are doing to make it work.
I say all of the above and the reality is that most teams don’t seem to have as many issues with it but I do see a number of problems every year that just come down to NT gremlins and I’m kinda not ok with that - which is the bulk of my frustration.
Thanks!
We also used the Limelight in 2018 for tracking cubes on the ground. We didn’t realize that it was so good it would track cubes that were on the ground, and also in the switch. So its utility was diminished on a real field, somewhat.
In 2017, I wish we had a vision tracking system to track gears on the ground. In conjunction with spatial awareness (SLAM-lite), it would have been really nice to have a pseudo-mapping of gears on the field as we drove around. That way we could have easily recognized gears positioned against our own driver’s station wall. For that particular use case, the Limelight would have excelled.
From a mechanical perspective, can anyone comment on its ruggedness? Did anyone at some point feel like this was an $800 investment instead of $400 (to buy a spare)?
Lots of people did buy two but every single one of them I’ve held/touched/attached to a Zebracorn robot, from the early betas until now has felt like a very mechanically reliable product.
It doesn’t feel cheap and it should stand up to most of what FRC can throw at it… though a direct impact from an aluminum portion of another robot at speed would likely break it… there isn’t much that is going to withstand that. It will definitely survive impacts from game pieces and other things. I think you could invest in just one.
We had the extra USB port (for second camera streaming) break on ours, so we use that one for testing.
As a general guideline (and not limelight specific), I wouldn’t want to walk into an event without a spare of a key component.
On this issue, I am in 100% agreement with Marshall. NT may have a place in FRC, but there are alternatives, JSON for instance, that just work better and are very well supported in multiple languages.
While I can not speak specifically to the ruggedness of LL, I think Marshall did a fine job of that above. I can say that the JeVois** is not** as rugged as the LL.
While competing in N. Arizona last year, the scenario Marshall described happened to our JeVois. Let’s just say it was a good thing we had a spare already configured and calibrated. It was only a $50 loss, but a loss none the less.
This is actually one of the primary reasons we like the JeVois so much. At $50 a piece it is a lot easier to buy a few more for development, spares, practice robot, kids to take home and tinker, etc.
As for durability, we have not lost one in competition yet, but I did have one usb port get wrecked at the hands of team members – so certainly not indestructible, but actually here again being cheap enough that the one on the robot, stays on the robot (even if I have to keep reminding some of our team that this is the case) is a big advantage.
As long as you mount it with the precautions you would typically take when mounting a camera (IE, don’t put it somewhere that will be frequently smashed into by things), then I don’t see there being any problems. We had ours mounted to our elevator/arm (near the center of the robot) this year and never had any issues with it.
Please do yourself a favor and listen to this advice.
We bought 3 limelights for 2018, and had no reliability issues with any of them.
Only downside to the limelight is if you don’t have a network switch on your robot you can’t do validation of it / see it while tethered via USB.
At a competition, are the only options to either install a network switch on the robot or unplug the limelight if you are needing to deploy code to your robot or test your robot at the practice field?
You could deploy your code over USB?
As marshall said, you can still deploy / tether to the robot over USB, but you can’t view anything from the limelight.
This year we normally deployed code by unplugging the limelight from the radio and plugging in our development machine.
If we had any issues / complaints about the limelight and needed to test it in the pit, we had a switch nearby that we could plug in the limelight, radio, and driver station / development machine as well.