Today we have the pleasure of welcoming Professor Miroslav Krstic from the University of San

Diego with us.

He is visiting us in the context of a grant that BacaTech, the agency, the association

of Bayer for the cooperation of California is providing to us.

As you all know, Professor Miroslav Krstic was born in Serbia.

Then after his bachelor and master in Belgrade, he moved to the U.S. where he got a PhD in

I think Santa Barbara, California under the advice of the celebrated Professor Kokotovic.

Then he got a position in Maryland.

Then soon after he came to San Diego where he has been running a position for maybe over

20 years where he has done everything.

He's nowadays a distinguished professor.

He's also a deputy for the vice chancellor of research, I think, of the university for

the last 10 years.

So those that know Miroslav and even those that don't know Miroslav, they know his name.

He's associated, you know, you can easily identify his name with many of the most, say,

dynamical and contemporary topics in control theory, in particular, backstepping, stabilization

and control that he has been developing for the last 20 years.

And he was, you know, due to COVID, his visit was to be delayed by, say, roughly two years.

So eventually taking the opportunity that he's in Europe, we thought that it would be

a great opportunity to inaugurate the lecture series of MOB, the recently launched Research

Center Mathematics of DERA of Frederiks-Aleksandr University.

And well, for us, it's really a great pleasure and great honor to have Miroslav with us.

Thank you for all that, you know, we're willing to come here and especially also to those

that are following us from abroad in Zoom.

Hopefully things will work and you will enjoy a lecture that I'm sure will be extremely,

say, dynamic and passionate in on the contemporary, say, interest of Professor Christy of MEDEOS.

Thank you very much, Enrique, for this introduction and for having me here.

Good afternoon, the Europeans.

Good morning, both the East Coast and my West Coasters in the United States.

It's really great to be here in Erlangen now for the second time after a visit six years

ago to Professor Leugering.

What we've experienced in 2019 after Professor Enrique Zuozua moved here is a veritable Zuozua

miracle.

It's not just the activities of this chair that are impressive, truly a tsunami of activities

by various people and connections, but also his own development beyond the traditional

PD control and related problems into terpite problems, deep learning, their various connections.

I'm so impressed by what's happening here.

You can look it up online.

I'm not quite ready to speak about machine learning properly, even though this is the

subject of our Bakheteq collaboration.

Their interesting developments they introduced in an informal meeting today about the use

of machine learning representations in neural networks in adaptive observers.

I've chosen for today a topic that is more solid, more mature, a topic related to my

long-term interest in extremum seeking.

There is a connection.

These are two distinct forms of learning.

The distinction is that machine learning conventionally attempts to learn a map, whereas extremum

seeking aims at learning only the optimizer of a map that is at least locally convex.

This is what makes extremum seeking usable in real-time applications for dynamical systems,

which has been my occupation the last couple of decades.