What a fun project! I wonder if that 'good sensor' would be good enough for pedestrian-carried dead reckoning navigation? Normally this is thought to contain so much noise as to be impractical.
A really good attempt, unfortunately the physics of balancing a bike moving in a straight line and one making a turn are vastly different and it would be really difficult to implement an automated way to make the turn. Not only does the bike need to momentarily steer in the opposite direction to initiate the turn it also needs to lean into the turn and not have the gyros try and stand it upright, but they also can't just hold the bike at an angle because the precise angle required varies based on speed and turn radius and if any of it is off the front wheel is just going to lose traction and the bike will go down. The small RC bikes with gyros don't even move them actively, they're just fixed in place and the bike just sort of does it's own thing as it leans and turns.
This reminds me of when I was a kid. We were nowhere near as technologically advanced, but anytime we had a hair-brained idea that didn't turn out the way we planned, it just turned into an excuse to epicly destroy something which always made us smile. Thanks
@We.Make.Machines that is exactly how I approach my failures these days. But my wasted youth did not afford me the luxury of forethought. Most of our epic destructions involved fire!
I think this could be a magnetic field related issue, which would explain that it can't be treated with thin foil. Just a theory: Maybe the tiny mass inside the MEMS accelerometer has a large enough µ_r such that it got slightly attracted towards the motors.
@@SirEngelmann definitely possible. It could balance indefinitely with the gyro motors spinning but maybe the 13hp motor emf had an effect. Something to test next time!
Could you use a hanging weight pendulum with a rotation sensor on the axel as your angle sensor? As a bonus, it would automatically lean in the correct direction during turns.
I'm the opposite of an engineer in my opinion, but I would guess the center of gravity and the center of mass might not be ideal :D Thanks for sharing the build process though. Even though you managed to remove the human from the bike, the humans doing the engineering still caused issues. Even a proxy won't help in this case hahaha. But I enjoyed watching this and I hope you guys are willing to give it another try! Oh, and ask James Bruton, he can 100% help you out with tips on how to make this work!
Just out of curiosity, when you turned the handlebars at speed , did you turn into or away from the direction of turn? I ask because that kind of looked like it might've cause the crashes. Nice job though, and now I have to look up LQR.
Rather than deal with sensors and code I wonder if you could have used ideas from the self-balancing monorail (th-cam.com/video/kUYzuAJeg3M/w-d-xo.html) and used the natural precession of the gryos as an input.
At the University of Märlardalen they managed to make a bike drive at its own with out fly wheels by simply turning in to the direction it falls! In the video below it don't carry a doll but I herd that they tested that to! th-cam.com/video/vICDzgeLqlM/w-d-xo.htmlsi=nj5O_fFkOF7CA_zA Anyway I like this channel. Keep it going, you rock!
A few points... To make something balance, you need to counteract any undesirable motion with more counter motion than is present. That is, if angular velocity is at 1, you need -1.01 to -2 to counteract it. The amount you put in depends on three factors. How far off your target you are, and how soon you will hit your target, and how responsive you want your balancing system to be. If you get oscillations, your algorithm is not doing a good job. And if you overshoot (which is what oscillations are) you also might have set it to be too responsive. In that it tries too hard to balance itself when the mass it's trying to balance is lower than the force you are putting in. A properly implemented balancing algorithm will self adjust such that it doesn't overshoot more than a few times, it will know that it's too good for it's current load. A standing bicycle is not al all the same as a rolling bicycle. Most bicycle geometries are made to self right themselves when rolling, which is why you can push a bike and it will stay upright for a while after you let it go. If you have a balancing algorithm that fights this self balancing, you are going to end up making things worse. A sensor that doesn't know up from down is a bad choice regardless of its quality. It would have been much easier to use physics to your advantage and just have a pole that controls some actuators. Not electronics. To make a bicycle turn, all you have to do is lean it. You don't need a servo on the handle bars. And an electronic angle sensor is going to mess you up no matter what you do because it won't know how much to lean to achieve balance in a turn. But a mechanical pole balance will know, because its reference follows the bicycle.
Thank you for watching and please consider joining our Patreon to help bring more crazy ideas to life and inspire future engineers!
You made this a lot more difficult than it needed to be. Which makes it absolutely exquisite content. Bravo.
Glad you enjoyed it! 😂
6:56 Thank you for reminding me of the last video. I got caught up in watching before I could recall which channel it was I had subscribed to lol
What a fun project! I wonder if that 'good sensor' would be good enough for pedestrian-carried dead reckoning navigation? Normally this is thought to contain so much noise as to be impractical.
Thank you for watching!
I just about panicked when he reached to catch the falling bike around the spinning wheel, that's a great way to lose fingers!
Evan had to risk it for the TH-cam algorithm!
A really good attempt, unfortunately the physics of balancing a bike moving in a straight line and one making a turn are vastly different and it would be really difficult to implement an automated way to make the turn. Not only does the bike need to momentarily steer in the opposite direction to initiate the turn it also needs to lean into the turn and not have the gyros try and stand it upright, but they also can't just hold the bike at an angle because the precise angle required varies based on speed and turn radius and if any of it is off the front wheel is just going to lose traction and the bike will go down. The small RC bikes with gyros don't even move them actively, they're just fixed in place and the bike just sort of does it's own thing as it leans and turns.
excellent
seems like a good way to mangle a low weight bike to to put big heavy gyros on it
You know the funny thing. If you just got a bike and slapped a motor on it and made it go fast it would keep itself upright without a gyroscope
But how do you go from 0 to fast before it falls? Maybe we try rocket engines next time?!
@We.Make.Machines push it down a hill, I've done that before and it stayed up
Yeah they over complicated that one
This reminds me of when I was a kid. We were nowhere near as technologically advanced, but anytime we had a hair-brained idea that didn't turn out the way we planned, it just turned into an excuse to epicly destroy something which always made us smile. Thanks
As the saying goes, when life gives you lemons, document the failure and violently disassemble everything so you can use parts for the next project.
@We.Make.Machines that is exactly how I approach my failures these days. But my wasted youth did not afford me the luxury of forethought. Most of our epic destructions involved fire!
Such a good company police. Very human.
Only the best at wmm
pretty sure the EMF produced my your motor woukd interfere with the sensor so, try shielding the sensors with basic protection.
Good thinking. We did try grounded foil but certainly could have tried other ideas.
I think this could be a magnetic field related issue, which would explain that it can't be treated with thin foil. Just a theory: Maybe the tiny mass inside the MEMS accelerometer has a large enough µ_r such that it got slightly attracted towards the motors.
@@SirEngelmann definitely possible. It could balance indefinitely with the gyro motors spinning but maybe the 13hp motor emf had an effect. Something to test next time!
@0:01 🙄 I'm not stupid, it's just the helium balloon holding it up! 😁
😂 busted!
Could you use a hanging weight pendulum with a rotation sensor on the axel as your angle sensor? As a bonus, it would automatically lean in the correct direction during turns.
I'm the opposite of an engineer in my opinion, but I would guess the center of gravity and the center of mass might not be ideal :D Thanks for sharing the build process though. Even though you managed to remove the human from the bike, the humans doing the engineering still caused issues. Even a proxy won't help in this case hahaha. But I enjoyed watching this and I hope you guys are willing to give it another try! Oh, and ask James Bruton, he can 100% help you out with tips on how to make this work!
This is really awesome!
Thank you for watching!
Just out of curiosity, when you turned the handlebars at speed , did you turn into or away from the direction of turn? I ask because that kind of looked like it might've cause the crashes. Nice job though, and now I have to look up LQR.
Sensor integration is hell, ask me how I know.
10:35 Is this a American Psycho reference? 😆
😜
Was looking for this comment... Oh my God it even has a watermark!
Rather than deal with sensors and code I wonder if you could have used ideas from the self-balancing monorail (th-cam.com/video/kUYzuAJeg3M/w-d-xo.html) and used the natural precession of the gryos as an input.
At the University of Märlardalen they managed to make a bike drive at its own with out fly wheels by simply turning in to the direction it falls!
In the video below it don't carry a doll but I herd that they tested that to!
th-cam.com/video/vICDzgeLqlM/w-d-xo.htmlsi=nj5O_fFkOF7CA_zA
Anyway I like this channel. Keep it going, you rock!
Did you use kalman filter on the "bad" sensor?
We sure did. The sensor would recover eventually but it took minutes even after optimizing parameters.
@@We.Make.Machines I see, btw Great Job on the build! Cant wait to see more from you guys! Love the comedy, love the effort, love the quality. Amazing
Billy, NO!
😢
NJ Drones explained
haha
how do you guys have so much machining and engineering knowledge but you cant figure out how to calculate the length of a chain 💀
A few points...
To make something balance, you need to counteract any undesirable motion with more counter motion than is present.
That is, if angular velocity is at 1, you need -1.01 to -2 to counteract it. The amount you put in depends on three factors.
How far off your target you are, and how soon you will hit your target, and how responsive you want your balancing system to be.
If you get oscillations, your algorithm is not doing a good job. And if you overshoot (which is what oscillations are) you also might have set it to be too responsive. In that it tries too hard to balance itself when the mass it's trying to balance is lower than the force you are putting in.
A properly implemented balancing algorithm will self adjust such that it doesn't overshoot more than a few times, it will know that it's too good for it's current load.
A standing bicycle is not al all the same as a rolling bicycle. Most bicycle geometries are made to self right themselves when rolling, which is why you can push a bike and it will stay upright for a while after you let it go. If you have a balancing algorithm that fights this self balancing, you are going to end up making things worse.
A sensor that doesn't know up from down is a bad choice regardless of its quality. It would have been much easier to use physics to your advantage and just have a pole that controls some actuators. Not electronics.
To make a bicycle turn, all you have to do is lean it. You don't need a servo on the handle bars. And an electronic angle sensor is going to mess you up no matter what you do because it won't know how much to lean to achieve balance in a turn. But a mechanical pole balance will know, because its reference follows the bicycle.