Getting ready for ECSg meeting

[From Bill Powers (980526.1625 MDT)]

I'm blowing off the last couple of days' messages -- they just call for
speculation anyway, so nothing I would say would be important.

I'm happy to report that the inverted-pendulum control system simulation is
working like a dream, somewhat to my surprise. It won't yet bring a
pendulum from the dangling position to a balance in the upright position,
but I've been looking at other people's approaches (do a Yahoo search on
<inverted pendulum control>) and I don't believe they do it either. What
mine does do is let you use the mouse to set a reference position for the
bob, and the cart moves to put the bob there, very quickly and without
overshoot or unnecessary undershoot. The "official" method is unbelievably
complex and nonintuitive, and it doesn't work real well, either.

The writeup for the conference is done, and it looks as though I'll have
copies of the paper to hand out. I'll post everything when I get back.

Best,

Bill P.

[From Bruce Abbott (980526.2055 EST)]

Bill Powers (980526.1625 MDT) --

I'm blowing off the last couple of days' messages --

All three of them?

they just call for
speculation anyway, so nothing I would say would be important.

I wouldn't defend the "universal error curve" idea either, if I were you.

I'm happy to report that the inverted-pendulum control system simulation is
working like a dream, somewhat to my surprise. It won't yet bring a
pendulum from the dangling position to a balance in the upright position,
but I've been looking at other people's approaches (do a Yahoo search on
<inverted pendulum control>) and I don't believe they do it either. What

See http://www.quanser.com/products/ip_se.htm

They do it.

What approaches do "other people" use?

mine does do is let you use the mouse to set a reference position for the
bob, and the cart moves to put the bob there, very quickly and without
overshoot or unnecessary undershoot.

Sounds great! If I get some free time I'd like to try this one myself, so
please don't give away your solution just yet.

The "official" method is unbelievably

complex and nonintuitive, and it doesn't work real well, either.

Please describe the "official" method.

The writeup for the conference is done, and it looks as though I'll have
copies of the paper to hand out.

Your prep work is done, so you now have plenty of time to respond, right? (:->

Regards,

Bruce

[From Bruce Abbott (980526.2055 EST)]

I don't. I can see what's arbitrary about it, and what is useful. No big deal.

I'm happy to report that the inverted-pendulum control system simulation is
working like a dream, somewhat to my surprise. It won't yet bring a
pendulum from the dangling position to a balance in the upright position,
but I've been looking at other people's approaches (do a Yahoo search on
<inverted pendulum control>) and I don't believe they do it either. What

See http://www.quanser.com/products/ip_se.htm

They do it.

I've downloaded some of quansar's movies; they look great. Notice that they
refer to "classical" control theory -- that's the kind we use in PCT. The
other types are what I referred to as the "official" control theory of today.

I'm going to have to try the double inverted pendulum. After looking at the
movie I can see how they do it -- the lower pendulum has to have a much
shorter period than the upper one. I notice that Quansar's system
oscillates continuously.

The self-erecting pendulum uses two modes (I deduce from the movie). One
swings the pendulum until it's just vertical; then the control mode kicks in.

What approaches do "other people" use?

Do a search on <inverted pendulum control> in Yahoo and have a look at some
of the first few links to see what is being taught.

mine does do is let you use the mouse to set a reference position for the
bob, and the cart moves to put the bob there, very quickly and without
overshoot or unnecessary undershoot.

Sounds great! If I get some free time I'd like to try this one myself, so
please don't give away your solution just yet.

OK, I won't. You have about two weeks.

The "official" method is unbelievably

complex and nonintuitive, and it doesn't work real well, either.

Please describe the "official" method.

All I can say is that it involves very complicated matrix mathematics and
lots of trig functions and is over my head. See the Yahoo reference above.

Best,

Bill P.

ยทยทยท

At 08:55 PM 5/26/98 -0400, you wrote: