Thursday, April 28, 2005

Another Seminar: Andy Teel, today

Andy Teel came to OSU today. He is a huge name in nonlinear control and has contributed an enormous amount over the last 15 years. It was a big deal that Professor Serrani convinced him to come and speak for us.

It was a good lecture. The topic was "Hybrid Dynamical Systems and Nonlinear Feedback Control," so he talked about these Lyapunov-like aggregate functions we can use in nonlinear systems that have both continuous and discrete counterparts. One good example was a bouncing ball -- it's continuous in the air, but every time it touches the ground it becomes discrete.

He used the development in this area to change how to approach some more classical problems. In particular, he addressed the problem of swinging up a pendulum so that it balances up vertically on a sliding cart that ends up in the same position it started in. This was fun. However, he had a really fun example dealing with HAART treatment of HIV patients.

One of his doctoral students was interested in this application, so they dived in. You see, (I'm going to screw this up) the HAART treatment reduces the number of infected cells. However, the body manufacturs Killer-T cells at a rate proportional to the product of the number of infected cells and the number of Helper T cells. So while you want to treat the patient, it would be nice to convince the body to produce more Killer-T cells so that the patient could be taken off treatment.

This, at first, suggests an open-loop control method, and this method has been tried and has some success with models. If you get the TIMING just right, you can take someone off treatment for a short period of time, allow the infected cells to accumulate, and the body will start producing large numbers of Killer-T cells. Then you start treatment again and the Killer-T's combined with the HAART treatment are very effective at lowering infected cells. Eventually, you remove the treatment completely and the Killer-T's keep the patient from needing treatment again. You haven't cured HIV, but you've treated it to the point where the body can take over.

However, this open-loop method isn't very robust. You simply don't know what timing is right for each patient.

So you "sample" each patient's blood every week and treat this as a hybrid dynamical system. And you know what? This feedback system was effective with up to 100% of sampling error (there's a lot of noise with those blood samples) and up to 15% of parameter error (as in, different patients, etc.). It was pretty sweet. I wonder if anyone's actually trying this? It was great to see controls applied this way.

And on top of this, Teel is just a regular American guy. It's not common to see a regular American guy giving an advanced engineering seminar. Kinda made me ... hopeful.

So that was fun today. I enjoyed it much.

6 comments:

Ted Pavlic said...

I actually found one of his old presentations, which includes some of the slides he used today:

A short course on hybrid systems and control: Lecture 1 - overview

Ted Pavlic said...

Oh, and after the seminar I helped a friend of mine carry some hardware up to his laboratory, and on the elevator ride up, we rode with Teel (and my adviser). So that was exciting.

Not as exciting as riding in an elevator with, say, Brad Pitt, but it's up there somewhere.

Anonymous said...

Brad Pitt would be the culmination of excitement for you?

Ted Pavlic said...

No, not the culmination, just an example. I was just showing that there exists at least one person other than Andy Teel who would be more exciting to have on the same elevator ride.

In effect, I wasn't showing that Brad Pitt is the culmination of excitement for me. I was showing that Andy Teel is *NOT* the culmination of excitement.

You see, it's a logic puzzle. You wouldn't understand. :-p

Anonymous said...

Your logic puzzle aside:
Was just checking to make sure Teel didn't wear chaps for his lecture..
Though, in my logic, that would put him above Brad Pitt's excitement influencing level. Now, Edward Norton would win whether or not someone else was wearing chaps. (He would win whether or not Brad Pitt was wearning nothing but chaps..)
But you wouldn't understand :-p

Anonymous said...

After further contemplation I would like to retract my previous statement. If I was in an elevator with a fully clothed Norton and Pitt in nothing but chaps I have to say that Brad Pitt would get me more excited. Hell, it would take Newton to beat that (and not the dead decaying Newton- a young Newton also wearing nothing but chaps).