The Poveda Lab Explains Control Theory
Watch this video to see researchers in the lab of Jorge Poveda, UC San Diego Electrical and Computer Engineering Professor, explain control theory as well as its real-world applications. In addition, the lab offers a peek into its ongoing work in the field of controls.
Learn more about how the Poveda Lab advances the field of controls.
Transcript
00:00:00 [Music] Control theory is the science that studies the role of feedback in dynamic systems. And when I say dynamic systems, usually what I mean is systems that move. They can be mechanical systems, but also electric systems, biological systems, even social systems. So what we study is how the, you know, using feedback, how using algorithms we can influence the performance and the behavior of these systems. So for example, how can we control a car to go from point A to point B while avoiding an obstacle? Or how can we improve the performance of power grids, for example, under disturbances? Fundamental sciences in general are about understanding how physical systems behave. Controls is about how we can influence that behavior of these physical systems. In that very broad sense, control theory is really almost everywhere. For instance, the very simple example of the elevator. On its own, the elevator is not going to go against gravity.
00:01:03 And it may go against gravity if we're putting too much force into it, for example. It will be jerky, it wouldn't be a safe, a safe system to operate. And so this is where control theory comes in. It helps us shape the behavior of this elevator so that it's safe for us to use. It's not as jerky, it's comfortable for us to ride in it and, you know, go upstairs. So control theory, even though we say that it's an invention of human intellect, feedback is one of the basic principles in biological systems. It's naturally occurring in the sense that, the systems that are naturally occurring, you can introduce some control action. How can I introduce another species that is going to control the population of, for example, rabbits in the field? So it's ubiquitous right now in our technological world and you can find it in the cruise control of your cars or even planes. More recently I've seen it being included for example in multimedia devices when you want to stabilize
00:02:01 the position of a camera when you're taking that shot. My focus is on designing algorithms that have two main areas of action. First one is I want to ensure that whatever we're designing is robust. And the other one is that we have good performance. And robustness is of utmost importance because, for example, if you want to implement something that is going to go inside of a car, you would like to have those control systems to be robust against arbitrary disturbances that you didn't model in your lab. So we are especially interested in something that we call hybrid dynamical systems. These are systems where you have interactions between the digital world and the physical world. Whenever you interconnect a computer with a physical system, there you got a hybrid system. We need to be able to understand how to model these systems, how to control these systems, how to optimize these systems. So that's what we do. We try to
00:02:52 design controllers and we try to understand also how these hybrid systems behave, how they should be modelled to achieve particular tasks. And we're especially interested in something called adaptive control, or adaptive systems, where the goal is to induce adaptive behaviors. So we would like these hybrid systems to be able to react without external inputs from humans to changes in the environment, for example, or to disturbances. So if there is some fault in a power grid, then we would like to be able to have an autonomous controller that, in real time and without external inputs from humans, is able to recover the system to some particular level. My main interests are in model-free control and optimization. Sometimes our models of the physical processes that we have are not accurate, or they involve some parameters that are not known to the engineer who is designing the controller that's supposed to influence the behavior of this physical system. And so,
00:03:54 here comes the importance of model-free control and optimization in the sense that we don't want to invest too much time in trying to model the underlying system. If there are unknown parameters or if there's the model is changing in time, then by interacting with the system in a clever way, we can extract information and then we can use that information in shaping the behavior of the system to our desired goals. So control theory is cool because I think it allows us first of all to have a good understanding of our world. You know, we operate under inputs and outputs, and then we, we react to events. And how do we react to events? You know that's essentially decision making. And it allows us to actually map those ideas into math. By leveraging these mathematical principles, then we can synthesize algorithms that you can go and implement in a computer, you know, in some hardware, and then see these algorithms in practice, you know, actuating on your
00:05:01 application of interest. And then you can see how math goes from equations to actual applications. And I think that's fascinating. UC San Diego has a really unique, offers a very unique environment to do research in control theory. There is a long tradition of control theorists here at UC San Diego. The fact that we have this interaction between multiple labs in different domains really gives our students the ability to collaborate with multiple professors, to learn different subjects, and to be exposed to the state of the art, essentially, in control theory, in a beautiful city like San Diego, where you can go to the beach, you know, any day a few blocks from here. I think that makes this a very special place.