Double inverted pendulum

[Orangebike666]

Can anyone explain this to me without insulting me or harassing me...?

[locomondoss]

Very good work

[HeNiNnG6]

wow...that makes me speechless 0.o

[hisanta32]

HOL-Y-SHIT... this is amazing!!! I'm honestly amazed, hahaa nice work man.

[super6plx]

that's amazing, surely eventually the upper half of that pendulum would just fold over and it wouldn't be able to compensate because if it goes one way it only stabilises the bottom half and the top half will still fall over, and if it goes the other way then the bottom half will fall over, thus the top one too. basically, I'm not sure how it keeps both halves of the pendulum straight when it should just fold and fall over

[cmayete]

@super6plx

You can cehck this video. It also withstands perturbations :-)

[jdwoods2008]

@super6plx

Think of someone backing up one of the trailer trucks with two trailers on it.

[super6plx]

@jdwoods2008 oh that makes sense I guess

[gidogirardi]

Hi,

I'm a brazilian control student, and my project is the same as yours. If you could help me, I would like to know what kind of connector did you used on the rotating axis of the first pendulum (which is attached to the cart), to get the signal of the second encoder (on the joint between the 2 encoders).

Thanks in advance

Vinicius

[angelrubiodiaz]

Hi,

There are no special connectors to the encoders. The cable was long enough to allow them to spin. I do not know if this answers your question.

Best regards,

Angel

[sovietspyguy]

Very cool! Hopefully we will get to do something like this in the motion control class I'm taking soon.

Thanks for the upload.

[ketamunke]

Not to be a dick, but why do you only show about 2 seconds of balance? Does it fail after it gets chaotic?

[angelrubiodiaz]

No, it stands still and it is able to stand in the upright position even if there are perturbations. There is another video that shows this.

[stuckintexas33]

how many encoders are you using? i am sure that you have one for the motor, one for the pendulum. im wondering if you have a third encoder on the second arm, i dont see a wire attached to it. i have a built an inverted pendulum using quanser and simulink but its only a single arm and i am interested in trying this double pendulum but i am trying to imagine the difference in the dynamics and how much my equations are going to change by adding another arm. my current system uses LQR to control it.

[angelrubiodiaz]

I am using 3 encoders. The cable is hidden in the second arm.

The dynamics equation are quite different from that of a single pendulum (and of course there is one additional DOF). If you want, we wrote these equations in the paper of the previous comment.

[controlprofessor]

Very impressive. I'm writing a paper that uses this type of system as an example. Do you have a report that can be cited?

[angelrubiodiaz]

We wrote a paper in IEE. The Title is "Swing-up control problem for a self-erecting double inverted pendulum". Authors: Rubi, Rubio, Avello.

IEE Control Theory and Applications

(pay attetntion: IEE instead of IEEE :-)

[antmosc]

hello, I'm an italian student in control engineering- I was looking for that paper but I couldn't find it in the internet.

can you help me? - thanks

[complacentsee]

hey, just wondering what system your using. Brand of system or brand of controller bored. and if your doing real time simulink. or dsp controller, just more info please.

I did this recently for a class. single pendulum, and the system was only capible of 8cm of total cart movement so we had to start from top. used real time simulink for the controller

[angelrubiodiaz]

The mechanical system was built from scratch. We used a commercial linear guide and we modified it developing a novel cable transmission. We used a Quanser board and real time simulink.

[controlprofessor]

I obtained and read paper. Your description of hardware indicates measurement of joint angles with shaft encoder. State variable control also requires velocities. How did you obtain them?

[angelrubiodiaz]

Perhaps, the answer is too technical to be placed here. Sorry for that to other users.

We develop a state variable control using the Ricatti equation. After the optimization, we know the location of the poles in closed loop. Finally, we got the velocities by differentiation of the angles and -to prevent problems form quantization noise- a low pass filter. We selected the cutoff frequncy 10 times larger than the freqs of the closed loop poles.

[rubenmalaga]

IS AMAZING!!!

Is the first time I write something here. I tried to make something similar with but working with one axis and one pendulum, but I didnt finish it. I never tought doing anything like this... For me looks like miracle.

I the kind of thing to shock your friends.

[Convert17]

HAHA! That's great. It's crazy that we can gte motors to not only balance those thigs upright, but be able to get them upright from a downward angle.

[angelrubiodiaz]

Especially if there is only one motor in the cart

[capra120]

Very cool and fast, check out our "Proyecto Equilibrista Electronica ITCR"

[roidroid]

is the cart the only motor, or is there also a motor inbetween the 2 pendulums?

[angelrubiodiaz]

There is only a motor in the cart. Between the pendulums there are no motors (just encoders to measure the relative angle).

[wtanaka]

impressive

[angelrubiodiaz]

Thanks!

[mistamustard]

Wow, you just optimized the joint's torques? You don't suppose we could use the same process to zero a pendulum with a given starting position, could we? If so, it'd take a bit of the chaos out of my life :)>.

[angelrubiodiaz]

Yes, you can. But since you do not have to generate a trajectory, it is easier to use "classical control theory" to take the pendulum to the down position.

[mistamustard]

Ah, sweet. I'll have to check that out and see if I can make it work in simulo.

[ATown31061]

wow, that'S amazing!! seems to besome kind of artificial intelligence!!

[angelrubiodiaz]

Not really. We performed a optimization: we assumed that there were three engines (one in the cart and one for each of the pendula). Afterwards, we performed a optimization to minimize the torques of the engines that do not exist and voila!, it worked.

[ATown31061]

Are you studying in a University in the usa?

[angelrubiodiaz]

No, I am a lecturer in a Spanish University (TECNUN).

[phlegmie]

frickin awesome, i'm doing an ip for my fyp, just single with digital control through a plc and its kinda kicking my ass!