11/19/2013 Jonathan Damery, ECE ILLINOIS
Written by Jonathan Damery, ECE ILLINOIS
One street-racing car is in close pursuit of another, wending, at various speeds, around a network of roads. For the tailing driver, the process requires a close coordination between observations and control, between the eyes, the hands on the steering wheel, and the feet on the pedals. If the car in front turns or changes speed unexpectedly, the other must follow. To automate that kind of decision-making process—to create an unmanned car that could perform the same task—the system of sensors and controllers would have to adjust for the unplanned, the stochastic, possibly the adversarial.
Of course, the automated car, described above, is just one application of dynamic and stochastic decision-making models, but it highlights the complexity of designing and understanding these systems—the theoretical puzzles that Başar has been solving since even before he joined the Illinois faculty in 1981.
When multiple decision makers are involved in a scenario, this is the basic premise of game theory (which is well known outside of academia for the semi-biographical movie A Beautiful Mind, about early game-theory pioneer John Nash). In the classic game-theory models—first introduced in the late 1940s and still taught today—there are a finite number of options (sometimes only two). Observing that the front car has slowed, the trailing car could essentially choose to go or stop; that’s it. The likelihood of those options could be predicted, but the models couldn’t be used to analyze higher-resolution reductions in speed.
So when Başar released his book Dynamic Noncooperative Game Theory in 1982, which comprehensively studied models for dynamic games, where decision makers could have an infinite number of options, the importance was readily apparent. In most applications, like the automated car, the system isn’t choosing between two options; rather, the options are limitless. The car could choose to go 60 miles per hour, instead of 61 or 59. It could turn one degree to the left, or two or three.
“In most engineering applications, you run into problems where your action set is an infinite set,” Başar said, and even beyond the realm of engineering, in economics, for example, the implications are just as important. “If you have to decide on the price of a commodity, then that price does not take discrete values, but will lie in some interval, for example, between a lower limit and an upper limit. Further, you make repeated decisions over time, based on information you acquire along the way.” In some ways, Başar has done for dynamic games what Nash did for static games. He developed a theoretical explanation for their operation and function.
His seminal book on that topic is now in its third edition, but a consummate researcher and writer, Başar has gone on to publish five additional books, the most recent of which, Stochastic Networked Control Systems, was released this year. He has authored over 250 articles in journals and edited volumes, and 400 peer-reviewed conference publications. He’s also the editor-and-chief of Automatica, which is rated with the highest impact factor in the control systems field.
Networks are the theme of his most recent research. Using a control-systems framework, he has, for example, proposed improved methodologies for network security. “When you play chess, there is one winner and one loser…Likewise, in security, if the intruder is successful, then it can disrupt the operation of the network, and that’s a win to his [count]. If you can prevent that from happening, then it’s a win for the network service provider,” Başar explained.
By thinking of security in this way—as a zero-sum game—it allows the service provider to design more appropriate safeguards. “Most of the measures taken were passive measures: building taller and taller walls for the castle,” Başar said of previous cyber security approaches, “but ours is more strategic, and you anticipate how the adversary will think and how the adversary will behave, within certain constraints, and then design your protection mechanisms accordingly. So that’s an anticipatory policy, strategy, rather than being a passive strategy.”
The IEEE Control Systems Award is in recognition of this indefatigable work, as were the other three awards he referenced earlier: awards given by the American Automatic Control Council (Bellman Award, 2006), the International Federation of Automatic Control (Quazza Medal, 2005), and the IEEE Control Systems Society (Bode Lecture Prize, 2004). Başar holds the Swanlund Endowed Chair at Illinois and is a professor at the Coordiated Science Laboratory, the Information Trust Institute, and the Center for Advanced Study. He will officially be presented with the IEEE Control Systems Award in December of 2014, at the IEEE Decision and Control Conference.
“Forty years,” Başar said, explaining the abundance of his research career and accomplishments. But “it’s not only [time], several of my publications are co-authored with my students. Definitely, we have quite a prolific and very talented group of students and post docs I had the privilege to work with, as well as faculty colleagues to interact with. And, this just keeps you going.”