7492 CavBots - 2022 Build Blog

Hi everyone! FRC 7492 CavBots, is excited to announce we will be joining the #OpenAlliance for the 2022 - Rapid React season, and hopefully future years!

For some backstory - we were started in 2019 in The Woodlands, Texas, a suburb just north of Houston. This will be our first year doing open design. We chose to start having a public build blog for a number of reasons - primarily, we think having that sort of documentation is beneficial long-term for both people in and out of the team. Plus, for families, friends, teachers, and sponsors, being able to follow the team makes it feel much more personal. Additionally, we believe that the community can give us direct help with our robot; by posting both the good and bad of our team, other people can learn from the good (and bad), and we can get help with the stuff we may be struggling with.

Good luck to all teams competing next season!

We will be competing at:

  • Channelview - March 11-13 (Week 2)
  • Amarillo - March 31 - April 2 (Week 5)
  • Hopefully FiT District Champs and World Champs!


Google Drive - will have all our public files, including CAD (will try to post a STEP file every few days, most likely won’t keep our original SolidWorks files updated, although we will make them available upon request). This will also have our media, so you can see our photos and videos as the season progresses.


Website - https://cavbots.wordpress.com/


To start off with, I’d like to give some background on our team in terms of capabilities going into the season. Our team started in 2019 with about 10 members, but has steadily grown to about 40 this season, which will be the largest yet. Always exciting! One of the main things to note is that we share a lab with team 1477, Texas Torque, and for the most part both teams pull members from the same school.

Some of the machines available to us are:

  • Probotix Nebula (3x4ft CNC router) - this is the workhorse of our lab. It cuts plastic very nicely, and also works well for aluminum. This accounts for the majority of our machined parts because it’s easy to use and has a large work area.
  • Forest Scientific CRP 2418 (18x24in CNC plasma table) - It has a small work area, but can hold tighter tolerances than we’ve seen with other plasma tables and of course is much, much faster for plate than the router. We use this extensively as well.
  • Tormach PCNC 1100 (9x18" CNC mill) - We don’t use this much because it has a small work area and a steeper learning curve than the other machines
  • Markforged Mark Two and Onyx One - These get used extensively. Aside from enabling us to print Onyx, the reliability and dimensional accuracy are handy in build season. We do have access to “normal” PLA/ABS/PETG printers as well, which is nice to save money.
  • Two lathes, one with DRO - We find ourselves not using lathes as much anymore. Often, it’s easier to buy standoffs or spacers premade off McMaster-Carr.

Kickoff day!
We spent a few hours and came up with a design. It’s VERY complicated, probably more complicated than it should be - but we also have plans to build a simpler version that can be made in a shorter period of time. Anyways, here’s what we came up with!

The balls with enter the robot through a full width intake, and then get dumped into a hopper. Then, they will go into a short vertical section, before finally going through the shooter, which will feature an adjustable hood and a turret. This is similar to what some teams did in 2020/2021, namely 4414 HighTide. Finally, there will be color sensors right before the balls enter the shooter, which can detect whether we picked up a blue or red ball and, if needed, reject the ball instead of shooting it in (probably by just changing the angle of the hood enough to miss intentionally).
The reason we chose to do this was because we felt that, by the nature of the balls falling back down after being scored, we believe there will always be plenty of balls in the relative center of the field to score. The adjustable hood and turret will allow the shooter to be spun up and aiming at the high goal. Or even low, if we so choose - could make more sense at longer ranges. Will have to prototype that. Anyways, since the shooter is “always ready”, we could leave the intake and indexer running constantly, so as soon as we “touch” the balls, they will get taken in, the hopper will center them, they will go up the center column, and then get shot. All (ideally) with some level of autonomy, allowing the drivers to focus purely on picking up balls.
The climb works by simply telescoping up and grabbing the bar, which gets us 6 points. We also toyed with the idea of having a second telescoping elevator on a pivot, which allows us to do a complicated series of handoffs to get to the 10 and 15 point bars. We haven’t ironed out the details on that, and we won’t until later in season when the drivers are practicing with the rest of the robot.
Thanks for reading! Let us know if you have any questions or concerns.


First major update of Build Season!
We’ve been designing out the robot, as you would expect. That started with a set of Master Sketches on day 1/2, which were of course iterated on as the days passed by. These are pictured below (The climbers are not included - they will sit on either side of the shooter column in the back).

These mainly allow us to start to work out systems integration, and more importantly, allow us to put together a general design, figuring out sensors, motors, and the geometry of the individual mechanisms, before breaking up into our respective subteams.

While working out all that, we also put together a schedule for the whole season, leading up to our first event in Week 2.

The important thing to note is driver practice; we set aside 2 weeks for them to have the robot effectively unrestricted. No matter what, they had to have it by then; even if we had to leave out the turret, climb, etc in the meantime. Besides that, the rest of the schedule is ambitious, having us have the robot mechanically done by mid week 3, however, we are confident it can be done.
This will be our first time really using a schedule like this - the idea is that we will talk about it and our progress at the start and end of every meeting, and try as a team to keep each other accountable to the schedule.

As for individual mechanisms, we prioritized the shooter, intake, and turret. Drivetrain design was already done from offseason and we were happy with where that was, and we determined shooter/turret to be important for drivers to have early, and the mechanisms requiring the most prototyping.

As for why we also prioritized turret, we had never attempted it before, and the entire robot was somewhat based around it. So, if we weren’t going to be able to realistically make it, we wanted to know right away so we could account for that. However, we did get something we are happy with, so we will move forward with it.

Here’s some pictures of our progress thus far on the intake and shooter

The intake is a 3 roller over-the-bumper intake, inspired by the many teams that did similar ones in 2020/2021. What is new, however, is the extra roller on the front; we are planning to experiment with some sort of “flapper roller” to catch bouncing cargo and direct it into the intake itself.

Here is the rough sketch and prototype of the shooter. 2021 bot for scale (and to power the motors). Because we have to shoot so flat, hence meaning very little shooter travel (under 20deg for a “fender shot”), we determined that we would need a “kicker tunnel” - something above the turret to accelerate the ball before it reached the shooter, allowing the shooter itself to mainly be used to aim the ball and also preserving it’s momentum, hopefully negating the need for an inertia wheel. This prototype exists primarily to test that theory, as well as to test hood and wheel material, and overall compression. We haven’t done enough prototyping yet, and I’ll post more as I learn more, but, with a grain of salt, what I have found so far is:

  • The “kicker tunnel” seems to work. It increased repeatability and exit velocity, although I won’t say it’s necessary since almost all our testing thus far has been while using it.
  • These balls compress quite a bit - up to and even over 2", even when fully pressurized
  • Slick wheels seem to struggle to get traction on the ball and slip a lot - we had to increase compression quite a bit to do that, enough that we were throwing off shots from them “popping” to size as they left the shooter, even when using the exact urethane drum we used in 2020/2021.
  • All our wheels were 4" - Colsons, AM HiGrips, and VEX VersaWheels all seemed to work well, while the aforementioned urethane drum was too slippery.
  • Ball pressure seems to matter a lot, and we were noticing wildly different results, while ball “fuzz” (we intentionally used velcro and made one of the balls extra fuzzy) does not seem to have a notable impact.

The main things we will be testing going forward are going to be focused on getting repeatable shots even with different pressure on balls, and then once we do that, we will also use this for testing angles and flywheel RPM for shooting from different positions on the field. I will keep y’all posted on our findings as we continue to prototype this in the coming days.

Finally, the turret. This will be belt driven, powered by a NEO on a 1:5 MaxPlanetary into 18:180 belt reduction, giving it a “travel time” of the entire ~330deg in around half a second. We planned to do a N550, but considering the structure we will need to put on top of it - essentially the entire shooter prototype pictured above save for the wood frame holding it up - we decided it was safer to just “overpower” it and put a full size NEO on it. We will also use the integrated NEO encoder to keep position, with a hardstop to zero it off.

If you want to look into this more, all STEP files are in the google drive I linked in the first post. Anyways, I think that mostly summarizes our progress so far! Let me know if y’all have any questions! Good luck, and thanks for reading!


Shorter post today. Was going to wait but I’ve seen a lot of positive feedback from when I sent this on the Open Alliance discord server, so I figured I’d share it here as well.

My team finished building the field!

It was…. So annoying. Frankly, we’re annoyed at FIRST for not getting a proper carpenter to make the drawings - for the most part, we just ignored them, and took dimensions off the CAD model and some actual carpenters (who happened to be parents of kids on the team) made it their own way. Regardless, we got it done, and importantly, were able to start running tests on it with the shooter prototype I mentioned in our most recent post.

Two back to back balls
Longish shot

Here’s what we found, of course, with a grain of salt. I would encourage everyone to do their own prototyping, but hopefully this helps with early design.

  • Ball wear (excessive “fuzz”, minor cuts and scrapes to the exterior, etc) were not a factor. We did not notice any difference in trajectory as a result of this.
  • Ball pressure made a huge difference. I don’t know exactly what they were at, but I know all of them conformed to the bounce description described in the manual. We estimated what we expect balls at events to be like, considering they can’t be expected to keep all balls at the same pressure. Anyways, ball pressure was so bad that despite extensive testing we never found a combination of ball compression and shooter speed and angle that was consistent across a range of pressures.
  • The shooter was reliable with balls of similar pressures
  • It’s actually pretty easy to hit shots as a Human Player. We had two people who spent a few hours (once the field was done) throwing balls at it and by the end they were already over 50% accurate.
  • Bounce-out was a big issue. As much as a third of the balls that made it in the goal bounced out.
  • Backspin made bounce-out worse - balls with high backspin (which was all of ours since we had a hooded shooter with no top roller) would seem to often get flung into the bottom of the cone and bounce out. However, balls thrown by hand, which had very little spin, did not have much issue
  • The “fender shot”, right up against the goal, was the most reliable at getting the balls in the goal across a range of hall pressures, but the worst at bounce-out
  • The shooter could shoot from almost anywhere on the field - by just varying the hood angle we could shoot from the fender all the way to the HP station with some level of accuracy - 50% ish? Didn’t do enough to get good numbers here

Moving forward, our plan is to do 1.5” compression with two 4” AndyMark HiGrips, powered by two NEOs on a 2:1 overdrive, and to experiment with adding a top roller. For other teams, I would honestly highly encourage building some sort of cone - it’s annoying, but really seemed to matter for shooter design.

Hope this helps. Let me know if you have any more questions or things you’d like us to try out at our next meeting on Tuesday.


Really wow, thanks Boomie. I’m super impressed and got some very useful information from this


I would be interested in seeing video of some of the backspin bounce out shots if you have some! Especially for up against the fender.


Done! Kind of. I spent a few hours today running tests, and uploaded videos of a good chunk of them. To start off with, I took the pyramid out and just screwed a wheel to the bottom - tomorrow I’m going to mount it and power it, likely very similar to how it will be on the real field (based on AM’s model and gear ratio).

Primarily, we shot from two places today: the fender, and what we’re calling the third zone. Basically, we have the field broken up into four rings, going from closest to farthest; the fender is Zone One (Z1), the colored rings around the hub where the balls start on (I forget the actual name of it, sorry) is Zone Two (Z2), Zone Four (Z4) is basically all the way up against the alliance walls/HP station, and Zone Three (Z3) is in between Z2 and Z4 and includes the protected zone against the climber. Not an exact science, and on the real robot we will have much more positional accuracy than that, but for just testing stuff we like this system.

We weren’t keeping count, but from my estimation we probably shot like 30ish times each from Z1 and Z3 today.

I’ll continue to post updates as I learn more and more. Let me know if there’s anything specific you’d like to see.

I uploaded all the videos I had to this folder in our Google Drive, but here’s a few in particular worth checking out (Red balls are close to correctly inflated, blue balls are partially deflated):

Zone One / Fender

  • For the fender it was a little higher than L3 overall, with most (~4/5) of shots going in, but a decent chunk bouncing out anyways. We found about 2/3 of the shots taken were successfully scored.
  • 3/4 hits
  • Hit, miss, miss, bounce-out

Zone Three

There’s quite a few more videos, I uploaded all the ones my team has taken to the folder, so if you want to look into the ball bounces in more detail I would invite you to go through and watch those.

EDIT: Fixed the links for the videos above. Sorry about that!


This is extremely helpful thank you.

This is also pretty scary.


This leaves a LOT up to the chains. I think teams that pay the money for Andymark’s upper hub will find that the it is well-spent.

EDIT: forgot about upper hub chains.

Chains? There are only chains in the lower hub right?


You’re absolutely right, you’ve caught me being a Tom Fool. This is worrying then.

1 Like

I’m almost amazed by how hard this is. Like, frankly, looking at 1678’s 5 ball auto, I think all of the last 3 would have bounced out. I have a hard time recommending any team aside from the very top (who wouldn’t listen to me anyways) do high goal… if your team is still on the fence this far in, and especially if you don’t have the space/money/materials/whatever to set up an accurate high goal, and above all, the time to prototype the heck out of it… you’ll have a very hard time making something that scores more points than just going low.


The thought of having frequent bounce-outs due to backspin is very scary indeed.

Does anyone have any thoughts on how the rotating agitator wheel in the bottom middle of the cone will affect bounce-out?

AndyMark is selling the full official upper hub with agitator on their website.

I’ll get the agitator installed tomorrow - I just stole the dimensions off AndyMark’s design and modified it to be coaxial which makes it easier to assemble. Won’t have it always running of course, plan is to have someone power it while we run some tests. I’ll let y’all know how it impacts bounce-out.


If the shot had no backspin, do you think / have data on whether it would bounce out less?

Hard to say for sure, but from shooting shots from the same places and similar arcs by hand, we were having much less bounce-out. In fact, shots thrown by hand hardly bounced out at all. I can’t say for sure that it’s because of the backspin, but from what I’m seeing, that’s my theory


I’m thinking Fender will be the money shot with PID and a consistent feed rate from a conveyor.

Your shooting documentation is some of the best out there currently. Really good work here, will be sharing with my team!


As one of the few teams with a realistic upper goal sharing information we really appreciate the videos you are posting. Would it be possible to capture some footage shooting from far, but also from the fender once you have the agitator installed. I am hoping we will have a more realistic goal built soon as well. Thanks.


Wow. This will be a lot harder than I thought it will be. Teams that don’t have the cone for the upper hub will be at a clear disadvantage when it comes to practicing their shots.

Thanks for posting!