You may want to watch part 1 first for context: ruclips.net/video/EEspocWIpu0/видео.html Final stretch before the robot competes. It drives well, flips somewhat well, and will no doubt be a learning experience. I've been squashing so many mechanical problems that they'll surely run out eventually, right? Plenty I had to omit to keep the video a reasonable length but may include that another time. Some immediate thoughts to get more power out of it; high quality AC clutches have a friction material inset around the perimeter maximise torque, and Cone clutches are worth exploring for their generally higher capacity (at a tradeoff of complexity). Didn't show any driving and flipping at the same time as I dont really have a safe space to do that, but from the short testing I did it appears to have practically no gyro effects. The flywheel gear grinds a little as the axle flexes under the load, so I just gotta always turn right :) Not as much robot footage as i'd like, the machine development itself has been very taxing which left little energy to set up a camera. Should get plenty from the event though.
I think your design is very cool, but I can't help but think the drivetrain of the flipper is overly complex. Why not ditch the whole rack drive and attach the clutch directly to the arm pivot with a roller chain? Or even just a connecting rod with rod ends. Like a crank rocker mechanism. You could still design this to vary the motion ratio if you think it's important.
@@zachary3777 That TO package was a diode and while the KSJ30D100-L probably contains MOSFETs, it is much bulkier and pricier, than designing a custom MOSFET based load switch PCB.
Making a dedicated, custom FET board would be the correct approach, but I appreciate the brutal reliability of that fat SSR for the moment. I'll consider optimizing to a custom solution in the future but right now I'm happy that's not causing me even more problems.
when fitting two pieces together like this, its normal practice to have the pin as cold as you can get it, and the hole as hot as you can get it before you start melting. this causes the hole to grow and be oversized. On the other hand causes the pin to be smaller. when the temperature between the two parts afterwards levels out, it becomes a very tight almost non removable fit with strength similar to if it had been welded.
Ideally you're not getting the parts as cold/hot as you can, but getting them to a target temperature difference determined by the dimensions of the parts and their respective coefficients of expansion. Too much heat or cold can damage a part in various ways, but also it's just nice to know if you're wasting your time with the freezer and/or blow torch and need to be removing material first.
@@BrokenLinkRobotics It doesn't need to be red hot. Put the shaft in the freezer for a few hours and get the gear to around 100c, you'll be surprised how easy it goes on.
That CAD overlay is such a great idea that I'm wondering why professional CNC machines don't do that. Shouldn't be too hard, add a (or multiple) small camera somewhere in the corner, a screen and some software.
The issue with that is normally (except for cases like Mazak's mazatrol), the only data the operator has at the machine is the GCODE that tells the machine where to position the tool tip, which contains no data about the finished part geometry. It also wouldn't do the operator much good to see the CAD geometry, as the operator is more concerned as to whether or not the tool is going to collide with their fixturing or exceed machine travel limits than what the final part is going to look like.
Most modern machines (and some old ones) including mine show the tool path rendered on a display. While it's not projected on the actual part, a marker is present where the tool is that reflects where the machine currently is so one can reference the model in real time.
For the pulley problem, perhaps you could try some sort of heavy duty motorcycle chain or Kevlar belt/ ribbon instead. Limit the amount of sidefumbleing it wants to do.
Good luck to you.... that system is amazingly complex and plain amazing- perhaps if you focus on mid power (instead of max) you can improve reliability but what an amazing job, just wow! Consistent flips to your back is often as good or better than big tosses in the air.
I really understand just how little sleep you've been getting now 😂 lookin good though and I see you've stepped up the editing with those overlays. I can't believe ai has actually taken my job (programming the disengage😅) best of luck for the event.
8:44 I know that sometimes you can heat the gear, and freeze the shaft. Gear slightly expands, shaft slightly shrinks. You have a small window of time where pressing is easier. Works better with a larger diameter, to the point where you can slip it on by hand, but this being smaller I think you would still need to use the press.
Love the "clickspring like" part previews before and during the cnc operations. If you want them to match better, you might need to use a renderer (like maya, nuke, keyshot, something with a virtual camera you can configure) to export the CAD previews. Most CAD programs will render with a set virtual camera, or in orthographic view, and that will never match the lens and sensor back plane of your IRL camera setup. Awesome robot and use of AI! Looking forward to see how the robot performs at competition. Good luck!!!
Try some loctite 648 bearing retainer as a substitute for a press fit like that. It has 4500 psi shear so pretty strong. On a 1 inch shaft with 1 inch wide gear that's a theoretical 589 lb-ft of torque.
That bot is an absolute work of art, and the video is really well produced too (others have mentioned it already, but I also appreciated the CAD overlays, which must have taken some work). Amazing perserverence too, to keep going after all those setbacks (I'm glad you show them, and don't just cut to the final working version, as if it's that easy). Would have liked to see a self-right attempt...
That looks incredible! Does the bot suffer from any bad gyro when turning while the flywheel is up to speed? Also thanks for supplying Team Panic with a replacement part for his key issues!
if it's like blip from battlebots it would yeah. there are potentially ways to mitigate it but those systems cost weight and design complexity, iirc the ripperoni team said the counter-flywheel didn't work with the (i forget) 12 or 30 pound one.
@@kaiserruhsam It doesn't need to have the same issues as Blip, their flywheel is mounted at a weird angle (almost like a very angled horizontal) which kinda amplifies the issue and causes the front to lift when turning one way while the flywheel in Loft is mounted much more traditionally like a vertical spinner which would make the sides try to lift up when turning. Lofts flywheel could also be running at lower RPM/stored energy hence making the gyroscopic precession not as severe. Just wondering if that'll become an issue since this is effectively a control bot.
I love the engineering of the mechanism, although or better because of all the obstacles to overcome. I hope, the streak of bad luck with small flaws like the controllers burning is not stopping you from solving the actual design problems :).
Printer threads can be tolerances using the XYComp feature, and with some clever design, the XYComp can be applied to _just_ the threads. Have been doing so for years with much success on many designs
May the driveshaft caps hold! Dang thing eh, almost need the caps to be bolts that screw in then grub locked or something. If it wouldn't interfere with the slot... Best luck!
Such a loverly looking bot, those machining overlays are pretty sweet Out of interest why didn’t you machine the gear first, then machine the shaft to the correct fit, I’ve always been taught that reducing diameter is always much easier to do and measure than increasing the size of a bore
It's like the Gell-Mann amnesia effect. ChatGPT managed to eventually, with enough prompting, accomplish a beginner programming task. The computer scientists and engineers behind ChatGPT did impressive work. ChatGPT did not.
Do you have any plans to change the design of the tip of the flipper? The current design doesn't seem super optimized for playing the ground game, especially when compared to the forks on Derive.
You're right, it's not great for ground game, but sportsman rules dictate limits on how much one can play. Will optimize if it ever competes in the open class.
I have it on good authority many third year engineering students in my cohort have reliably produced final project code with ChatGPT, predominantly in C++. It's a good tool in the right hands.
You may want to watch part 1 first for context: ruclips.net/video/EEspocWIpu0/видео.html
Final stretch before the robot competes. It drives well, flips somewhat well, and will no doubt be a learning experience. I've been squashing so many mechanical problems that they'll surely run out eventually, right? Plenty I had to omit to keep the video a reasonable length but may include that another time.
Some immediate thoughts to get more power out of it; high quality AC clutches have a friction material inset around the perimeter maximise torque, and Cone clutches are worth exploring for their generally higher capacity (at a tradeoff of complexity). Didn't show any driving and flipping at the same time as I dont really have a safe space to do that, but from the short testing I did it appears to have practically no gyro effects. The flywheel gear grinds a little as the axle flexes under the load, so I just gotta always turn right :)
Not as much robot footage as i'd like, the machine development itself has been very taxing which left little energy to set up a camera. Should get plenty from the event though.
I think your design is very cool, but I can't help but think the drivetrain of the flipper is overly complex. Why not ditch the whole rack drive and attach the clutch directly to the arm pivot with a roller chain?
Or even just a connecting rod with rod ends. Like a crank rocker mechanism. You could still design this to vary the motion ratio if you think it's important.
You should look into MOSFETs for the clutch. They can easily switch hundrets of amps. In fact, they are on pretty much every ESC you ever used.
@@VulpeculaJoy he did end up using one
@@zachary3777 That TO package was a diode and while the KSJ30D100-L probably contains MOSFETs, it is much bulkier and pricier, than designing a custom MOSFET based load switch PCB.
Making a dedicated, custom FET board would be the correct approach, but I appreciate the brutal reliability of that fat SSR for the moment. I'll consider optimizing to a custom solution in the future but right now I'm happy that's not causing me even more problems.
the cnc part overlay is pretty sweet
Agreed! First I've seen that, it adds a lot to the video. Would love to know how it was done
Those cad overlays are awesome
when fitting two pieces together like this, its normal practice to have the pin as cold as you can get it, and the hole as hot as you can get it before you start melting. this causes the hole to grow and be oversized. On the other hand causes the pin to be smaller. when the temperature between the two parts afterwards levels out, it becomes a very tight almost non removable fit with strength similar to if it had been welded.
beat me to it
Ideally you're not getting the parts as cold/hot as you can, but getting them to a target temperature difference determined by the dimensions of the parts and their respective coefficients of expansion. Too much heat or cold can damage a part in various ways, but also it's just nice to know if you're wasting your time with the freezer and/or blow torch and need to be removing material first.
Hmm thinking about it, heating the gear may not be an option. It has heat treated teeth and blasting it hot will degrade that.
@@BrokenLinkRobotics Liquid Nitrogen is often used for this, eliminating the need to heat the outer part.
@@BrokenLinkRobotics It doesn't need to be red hot. Put the shaft in the freezer for a few hours and get the gear to around 100c, you'll be surprised how easy it goes on.
I came here to praise the CNC overlay sequences, but I see I’m not the only one. Love your videos mate, I find them so interesting!
I love the Clickspring-overlays and Portal radio music!
I think the music is from TF2 MvM upgrade station, probably a cover of some sort.
UPD: yeah it is linked in the description.
@@JTCF You are right, my ears deceive me. I play tf2 I should know better lol.
That CAD overlay is such a great idea that I'm wondering why professional CNC machines don't do that. Shouldn't be too hard, add a (or multiple) small camera somewhere in the corner, a screen and some software.
The issue with that is normally (except for cases like Mazak's mazatrol), the only data the operator has at the machine is the GCODE that tells the machine where to position the tool tip, which contains no data about the finished part geometry. It also wouldn't do the operator much good to see the CAD geometry, as the operator is more concerned as to whether or not the tool is going to collide with their fixturing or exceed machine travel limits than what the final part is going to look like.
@@bmanc2000 It could show the toolpath instead
If holograms ever get to the point of syfi from fims
I want to project my toolpats in my macine to place my clamps around it
Visualizing the toolpaths would be so cool
Most modern machines (and some old ones) including mine show the tool path rendered on a display. While it's not projected on the actual part, a marker is present where the tool is that reflects where the machine currently is so one can reference the model in real time.
Use some grinding compound and lap the flywheel to the clutch. I used to repair AC clutches...
Hmmm interesting idea! Will consider it
Maybe replacing the rope actuating the flipper with some type of flat belt could reduce the sideloads and friction? I think that would squish less.
For the pulley problem, perhaps you could try some sort of heavy duty motorcycle chain or Kevlar belt/ ribbon instead. Limit the amount of sidefumbleing it wants to do.
Good luck to you.... that system is amazingly complex and plain amazing- perhaps if you focus on mid power (instead of max) you can improve reliability but what an amazing job, just wow! Consistent flips to your back is often as good or better than big tosses in the air.
Subscribed! The CAD overlays for the CNC remind me of Clickspring 💯
Amazing work, especially on the robot but on the video too.
That's really cool how you got the perspective to show the parts in the stock!
I really understand just how little sleep you've been getting now 😂 lookin good though and I see you've stepped up the editing with those overlays. I can't believe ai has actually taken my job (programming the disengage😅) best of luck for the event.
amazing progress dude, each iteration gets more impressive
These videos are amazing.
It would be great to have you come over to the states for NHRL.
8:44 I know that sometimes you can heat the gear, and freeze the shaft. Gear slightly expands, shaft slightly shrinks. You have a small window of time where pressing is easier. Works better with a larger diameter, to the point where you can slip it on by hand, but this being smaller I think you would still need to use the press.
Love the "clickspring like" part previews before and during the cnc operations.
If you want them to match better, you might need to use a renderer (like maya, nuke, keyshot, something with a virtual camera you can configure) to export the CAD previews. Most CAD programs will render with a set virtual camera, or in orthographic view, and that will never match the lens and sensor back plane of your IRL camera setup.
Awesome robot and use of AI! Looking forward to see how the robot performs at competition. Good luck!!!
Try some loctite 648 bearing retainer as a substitute for a press fit like that. It has 4500 psi shear so pretty strong. On a 1 inch shaft with 1 inch wide gear that's a theoretical 589 lb-ft of torque.
The overlays are awesome and something I haven't seen before. Genius !
Still can't name any more satisfying manufacturing montages than these!
Absolute madlad. I'd love to see your take on a melty.
interesting how the concept is similar to the robot blip but the approach in practice is still very different
I'm amazed it's working this well to be honest, it's a testiment to your perseverance and skill.
That bot is an absolute work of art, and the video is really well produced too (others have mentioned it already, but I also appreciated the CAD overlays, which must have taken some work). Amazing perserverence too, to keep going after all those setbacks (I'm glad you show them, and don't just cut to the final working version, as if it's that easy). Would have liked to see a self-right attempt...
nice combat robot, good cnc machined parts. very Shiny
In the UK and Europe that is classed as a "Class D Flammable Metal"
neat editing during the machining!
8:15 I got a dirt cheap set of telescopic bore gauges, I use them with mitutoyo micrometres and they work quite nicely.
Your robots and videos are so cool, keep it up!
Your robots are pretty neat too :)
That looks incredible!
Does the bot suffer from any bad gyro when turning while the flywheel is up to speed?
Also thanks for supplying Team Panic with a replacement part for his key issues!
if it's like blip from battlebots it would yeah. there are potentially ways to mitigate it but those systems cost weight and design complexity, iirc the ripperoni team said the counter-flywheel didn't work with the (i forget) 12 or 30 pound one.
@@kaiserruhsam It doesn't need to have the same issues as Blip, their flywheel is mounted at a weird angle (almost like a very angled horizontal) which kinda amplifies the issue and causes the front to lift when turning one way while the flywheel in Loft is mounted much more traditionally like a vertical spinner which would make the sides try to lift up when turning.
Lofts flywheel could also be running at lower RPM/stored energy hence making the gyroscopic precession not as severe.
Just wondering if that'll become an issue since this is effectively a control bot.
Gyro is minimal, the wheel itself is fairly light and doesn't spin very fast (relatively). Coupled with a wide frame it doesn't affect driving much.
Watching the receiver being "gently" sent off in silence broke me
You had me at flywheel flipper.
Good stuff! Keen to see how it performs
I love the engineering of the mechanism, although or better because of all the obstacles to overcome. I hope, the streak of bad luck with small flaws like the controllers burning is not stopping you from solving the actual design problems :).
Printer threads can be tolerances using the XYComp feature, and with some clever design, the XYComp can be applied to _just_ the threads.
Have been doing so for years with much success on many designs
overlaying the CAD drawings over the camera footage is brilliant and must have been a lot of work.
Intense and focus heat needed... *takes notes*
Hang on a second, that music at 1:17… is that… a remix of TF2’s upgrade shop music?!
Flippin awesome!
Good luck!
Just wanna say that I love your content and you are hella underrated ❤
Maybe putting some thrust bearing on the shaft before the hardox plates would help with loading?
This is so amazing it makes my featherweight look like a McD happy meal toy.
Can it self-right? Why not make a slo-mo of self righting?
Ah forgot to include those shots. Yet is can self right quite well.
this is really awesome
How much money does a bot like this cost to create
May the driveshaft caps hold! Dang thing eh, almost need the caps to be bolts that screw in then grub locked or something. If it wouldn't interfere with the slot... Best luck!
Nice overlay milling shots
Instead of a friction clutch... maybe an inductive one could be used?
Beautiful
i 100% agree with anyone who wants to see more doggie/machining clips :)
Such a loverly looking bot, those machining overlays are pretty sweet
Out of interest why didn’t you machine the gear first, then machine the shaft to the correct fit, I’ve always been taught that reducing diameter is always much easier to do and measure than increasing the size of a bore
The shaft was sent over by pcbway, so was predetermined. Good point though
Have you thought about springs or using capacitor's to store a huge amount of power and then dump it into a huge hub motor to spin you flipper.
So at 24:54 you are basically saying the best offensive weapon against Loft would be a lubricant spray?
Microphone sounds pretty bad today... Any chance for an upgrade?
Insane how impressive the AI can be in the right hands
It's like the Gell-Mann amnesia effect. ChatGPT managed to eventually, with enough prompting, accomplish a beginner programming task. The computer scientists and engineers behind ChatGPT did impressive work. ChatGPT did not.
so cool!
Love your content
Do you have any plans to change the design of the tip of the flipper? The current design doesn't seem super optimized for playing the ground game, especially when compared to the forks on Derive.
You're right, it's not great for ground game, but sportsman rules dictate limits on how much one can play. Will optimize if it ever competes in the open class.
Ltt screw driver?
Hey, nice video! What RPM are you running the flywheel at?
About 8000 I think? Hard to tell
Let's hope you can achieve some lofty goals as opposed to having to throw in the towel
What's your shop assistant (the dog's) name?
Hudson
What modeling software do you use
Fusion 360
@BrokenLinkRobotics thank you
To nestačí dolejt metabond?
Sorry man, that press fit bit should have you restart entirely with heated and cooled parts, that's just begging to slip on you
Poor derived 😭
"use ai for more involved software projects" HAH, ai is bad in general, but it fails horridly at anything more complex than primary school programming
I have it on good authority many third year engineering students in my cohort have reliably produced final project code with ChatGPT, predominantly in C++. It's a good tool in the right hands.
That music. 🤮