Side-Stepping of the Triple Pendulum on a Cart
Vložit
- čas přidán 18. 07. 2011
- Triple Pendulum on a Cart
Side-stepping
Two-degrees-of-freedom design:
Constrained feedforward based on the nonlinear model
vs.
Flatness-based feedforward control on the linearized model
with optimal feedback control
TU Vienna ACIN CDS www.acin.tuwien.ac.at/
Done by Andreas Eder and Tobias Glück
users.acin.tuwien.ac.at/tglueck/ - Věda a technologie
For non-experienced people who may not see how impressively precise those movements are, imagine backing up with three trailers!
It would be the equivalent of backing up with two trailers, since the truck body counts as one as well. Still unimaginably difficult and that’s without gravity
No I won’t
Hey I used to do that at work with a tug and 3 trailers. I would parallel park just the rear trailer so I could unhook it, then do the same with the other 2 in other places. You just have to be nice and slow and think about where everything was moving. I can't say it was easy, but it was doable. I don't think I could have done 4 but we weren't allowed to pull more than 3 anyway so I never got to test it out.
@@mishaberger1718 No, in your analogy the truck body is equivalent to the sled moving on the rail. Each rod in this system is equivalent to a trailer
I bet this controller has one hell of a transfer function :)
Very satisfying how the distance moved is exactly the wall panelling. Nnnggh. :)
Super impressive mathmatics I'd say.
Terminators will climb a mountain of skulls using similar advanced stability tech.
Hats off to you, this is really impressive.
This is very, very impressive.
Thanks, you've encouraged me to take up control engineering.
There are 56 transitons for a fixed cart position and for each one you have to find a suitable trajectory. The main problem is finding a trajectory that complies with cart position restrictions.
Does each joint have a position sensor or does the entire mechanism 'feel' the individual weights for each post? This is amazing stuff. Thanks for sharing.
very impressive!
For sure!
very impressive
mindblowing
You can tell how the linear model must make much more gradual movements so that the angle from the balanced position isn't too large.
This is impresssssive!! Did you think of addressing the twin inverted pendulum problem?
Awesome! How about the quadruple pendulum?
Are there sensors at the joints, or just at the base?
3 years later.... still wondering
There are rotary encoders at the joints to measure the angle.
Holy cow. That's all.
Do the segments need to be progressively longer to be controllable?
Now do this with a chain! :D
@Tobias Glük: Couldn't your system make the transitions itself between the stationary positions. Like going from "the 3 links" up to "only the red link down"?
I saw you did it for 2 links, why not 3?
encoder type you use it?
No, we use DSPACE.
I wonder what would happen if one were to manually interfere with it? There would probably be a certain point where it would have to "reset" before it could get back into position, because it can only go so far along the track
They've done that in other videos.
Why is that the non linear model causes it to move more erratically? intuitively I would think that the linearized model would result in more erratic movements. Appreciate any response.
i'm not sure but i think that the calculation time for the non linear model is longer than the linear model
I guess the "erratic" movements are optimal in some sense. It could be the result of minimization of the transition time given constrained cart position and velocity. Perhaps if energy, or total distance travelled, was chosen as optimization criterion, it could result in smoother trajectories.
But the model needs to be calibrated, right?
Does this use machine learning?
The computers are coming for us :(
Lol, pardon my ignorance.
If you can do three, then you can certainly try for four!!!!!!!
That is not necessary.
No, it uses magic.
Some say it is control engineering, but they are wrong!
No.
It failed
Now think about what religion brought us. :>