Self-Balancing Bot w/ 12V Parallax Wheels (attempt)
Got the motor mount and wheel kit from Parallax when saw it was half off:
https://www.parallax.com/product/motor-mount-wheel-kit-aluminum/
Also, got a caster to make a 3-wheel bot.
But, then saw a bunch of self-balancing bots and decided to try that first.
Here's the concept (see attached pic).
The construction is 1/2" HDPE and 1/4-20 hardware.
Just CNC'd the main board and it seems pretty solid, even with the giant hole in the middle for the battery.
How high up or how low to mount battery is perhaps the main question. Or, maybe if this battery is too big is another... Been reading that being top heavy is actually an advantage for self-balancing bots. But, think this SLA battery is beyond the pale, so lowering it via the giant hole. May try different amounts of lowering if this doesn't work.
Plan is to use P2 and BNO086 sensor in RVC mode to achieve balancing.
Might be a disaster, we'll see!
Comments
Hi,
"Steiner's theorem" https://en.wikipedia.org/wiki/Parallel_axis_theorem gives that the moment of inertia grows with the height in a quadratic way. As the moment of the weight corresponding to the deviation of a certain angle only grows linearly, it will be much more easy to balance, when you battery is mounted high. It is funny, that the long long rocket is a shape that is far more easy to balance than your shape. So if you succeed, the rocket will be easy-peasy for you next....
Christof
Hanno Sanders used cascaded PID
@Mickster I remember Hanno's bot from years ago...
Got the wheels assembled.
After looking at Arlo, thinking might be better to have motors on top. Mostly because of wiring...
The allowable rotation angle goes down a hair, but not much.
Ground clearance also about the same.
Pretty sure that bot is going to need to balanced to start out with. Need to add weight to the front...
That's the sort of thing that can be handled by an offset...added to the PID output. Same applies with vertical axes to compensate for gravity. I wouldn't needlessly increase the inertia...JMO
Think you can do an offset but then the platform wouldn't be level when balanced.
Funny that nobody seems to mention that anywhere, but pretty sure it's true.
Also read about needing to fuse in data from the accelerometer because rotation angle can drift over time.
Not seeing that with BNO086 in RVC mode though...
Seems rock solid.
BNO086 also seems to know about its absolute angular position, which is interesting.
It must be doing some kind of calibration at start up to figure that out.
Also read that one needs 300 RPM motors and enough torque.
These motors are rated at 95 RPM.
Thought that might be a problem for a second, but now thinking that the 300 RPM is for tiny robots.
This one is very massive and don't think it needs to move that fast.
Found an example on the web of a similar bot that worked.
Wasn't quite as massive as this one though, so think it good shape.
PID is wonderful. It can take care of everything. The offset can be adaptive.
One of the problems that I see with these hobby devices is low resolution feedback. You can't have too much.
I like those CUI modular encoders and 4096 line are available. This translates to 16384 quad counts/rev and we have Prop horsepower.
Oh heck, we are talking about servo motors, right, not steppers?
I not familiar with the product 😊
It’s just the parallax dc brushed motors from the Arlo bot
Going to use with parallax pwm MOSFET hbridge driver board
Hi, how did you model this? It would be great, if we could switch to FreeCad for all the mechanical designs. As we now see the version 1.0 https://freecad.org/index.php?lang=en
@ErNa FreeCad does look nice. Used SolidWorks for this because that's what we use at work...
That reminds me, I have to come up with an explanation why this is needed at work...
Main HDPE board was bowing a bit, so added a piece of t-track that was around.
Seems better now...
Liking the wheel supports. They certainly didn't cheap-out on this.
Found a note and code on using the Parallax Universal motor board with these wheels:
https://www.parallax.com/arlo-conversion-to-p2-with-the-universal-motor-controller/
Seems it should be on one of the product pages, but isn't....
Without having given this a fair shake, I shouldn't really critique it BUT "ramp-time". If ever there was a case for s-curve acceleration/deceleration, this is one of them. But the source is there, maybe we could/should give it a go(?).
It's all wired up. Moment of truth is near...
Hmm... Motors act a little different than expected...
They don't turn at all until drive set to 20% Doesn't turn with PWM set to 10%...
And, they make funny pulsing noises. That may have something to do with not having any load though.
They sound better with a little load...
Knee-jerk reaction is aforementioned lack of feedback resolution....but no clue what you have. Also if your PID I-term is >0 then set it to zero.
Observation is probably right. I'm not clear on the fundamentals around motors at all but I'd guess there is magnetisation losses of some sort that means you don't get torque until a certain amount of current is flowing. EDIT: Another factor can simply be mechanical frictions prevent motion until enough torque is applied. Gearboxes are a known source of this issue.
Roughness in commutation is also probably unavoidable to a certain degree. After all, the shape of the poles are not continuous. And for brushed DC you've also got the segmented transitions around the commutator.
PS: Also, be aware that voltage control behaves differently to current control. If you can do current sensing easily then I advise doing so. This is normally performed inside the motor drive circuit, with the demand from the motion controller (Propeller) representing desired current rather than voltage.
I have small Pittman (brushed) motors on my desktop simulator. They are smooth as silk and hold position absolutely with zero error. For me, the resolution is low @2000 quad counts/rev but even the lowest speed is totally smooth.
I don't think Rayman's motors have any encoders, and maybe no current sense in the drives either. They look like simple brushed DC, presumably with primitive voltage based speed control.
Looks like he's planning on relying on the BNO086 sensor for all feedback.
Oh, that would explain it. I'm only familiar with Hanno's Dance Bot. Wish he was still around. I used his Viewport software with P1 and would love to have a P2 version.
PID video with no yapping
I think we're gonna need a lot more resolution and some serious gain.
Sorta working...
I envy you guys that just seem to get things done
So your update from the sensor is 100Hz? I don't follow Spin too well.
Can you increase that update rate? I think I read that the sensor is good for 1KHz (?)
Going to try this dirt simple RVC (robot vacuum cleaner) mode first. It's just a serial output, fixed at 100 Hz.
Think it's close to working, just need to get rid of oscillations... Hopefully, fine tuning the PID will get there...