Thanks for the awesome resource again! The binder looks great as always.
thanks for sharing the technical binder.
That floor intake worked great in China.
254 Build Blogs will forever be the only paragraphs I willingly read at home.
Amazing resource, thank you for posting it once again!
On your drivetrain, what is the benefit to running chain all the way down the length of the chassis from the gearbox to the front wheels? As opposed to running chain from the gearbox to the middle wheel, and then from the middle wheel to the front wheel.
DISCLAIMER: I’m not on 254.
I would assume this is so that the breaking of a single chain doesn’t rob them of control of an entire side of their chassis.
We have two separate chain runs from the gearbox so that if a chain fails, only one wheel will lose power (as opposed to two.) Also, if we were to daisy-chain off the middle wheel, there would be double the tension load through the back chain, making it more prone to failure.
Disclaimer: I’m also not from 254.
I guess that another reason can be so that the backlash on the farthest away wheel is reduced since it’s the result of one chain run instead of two, improving handling or auton or something idk.
What’s the reasoning to switching to open-loop after the PID controller? Wouldn’t it be ideal to stay in closed-loop to maintain a constant RPM, especially if the number of balls in the shooter varies?
According to Kevin, they actually found much better disturbance rejection using open loop control. I’m still personally skeptical that modern control theory wouldn’t have provided quicker recovery if implemented correctly, but hey, it worked well enough to win champs.
Taking a guess, it’d probably be more reliable because it can’t saturate the controller. Also eliminates any control lag. Nevertheless, if 254 hadn’t successfully used it, I’d be skeptical of whether it would be helpful.
Can you explain how 254 calculates chain center to center distances, the sorts of fudge-factors that you add to those calculations, and the kind of tolerances that are acceptable during manufacturing to stay within acceptable chain tension?
The main issue we encountered was modelling the shooter’s closed loop control with balls interacting with 2 flywheels. Especially since our feeding system does not feed balls in a predicted manner, a number of scenarios can happen as balls make contact with the flywheels (2 balls can contact both flywheels at the same time, a ball can contact one flywheel while no ball is contacting the other, a ball can begin to contact one flywheel as a ball is leaving the other flywheel, etc). When our programmers were trying to tune a closed-loop control, because we did not know when and where balls were touching the shooter, any changes in the closed-loop control terms either increased or reduced the energy imparted onto the balls by undesired amount. Thus, we resorted to open-loop control.
In your tech binder, you have a graph vs time per cycle. But each strat has 2 “objects” or variables that you refer to. For example, strat 6 is “1 does gears, 1 does balls”. What are you referring to? I assume its not just your robot, but its also not an alliance because its only 2? Do you dedicate an entire robot as a defense bot?
Thanks in advance
When deciding our strats, we thought about either having it revolve around just our robot, our robot and another, and our robot and 2 other robots. Obviously, since in this game, any decent alliance of three robots would easily overpower an alliance consisting of just one really good robot, we decided to not plan our strategies around just our robot. We also assumed that, given our ability to shoot 40kpa, we would seed fairly high which would give us the chance to pick another good robot. We thought that all we needed in the third robot was for it to play a little bit of defense because if so, the opposing would not be able to get 4 rotors (in retrospect, we severely underestimated the importance of a third robot, especially when we saw the level of play at Champs). Thus, there would only be two robots capable of scoring and so we decided to plan out our strategies around two robots. Strats 1-8 only consider two robots. To see how much of a difference a third robot would make, we decided to include Strats 9-10 and as seen, even just the third robot’s ability to do a few gears greatly improves the score (we employed strategies similar to Strats 9-10 at Champs and FOC).
In short, the 2 objects/variables are 2 robots, one of which is us and one of which is our first pick.
That makes alot of sense, great strategy analysis!
How do you make so many prototypes?!?! :ahh: :ahh: :ahh:
With a laser cutter
What laser cutter are you using?
Saw you’re running vision off of a nexus? Is your app available on th google play store? If not would you be willing to release it to the public for other teams use?