The Decision and Control Laboratory (D&CL) at Illinois traces its origins to an “adaptive systems” research group established in the Coordinated Science Laboratory in 1959 by CSL associate director Mac Van Valkenburg. The group consisted of just two people: Van Valkenburg and his freshly graduated advisee, José Cruz (PhD 1959), whose dissertation extended traditional circuit theory to address systems with time-varying components, introducing ideas from the emerging field of automatic control.
Nearly 50 years later, the D&CL claims 20 faculty from five departments and enjoys a reputation second to none in the world of control, now a mature, highly interdisciplinary research field. Those who remember the early days of the Illinois group attribute its success to the style and philosophy of its founder, who was, perhaps ironically, anything but a control freak. “His philosophy was the same as [then ECE department head] Ed Jordan,” said Cruz of Van Valkenburg. “Hire the very best you can find, and leave them alone.”
Growing the Department
William Perkins finished his PhD at Stanford in 1961 and joined the group that year, initiating a productive collaboration with Cruz in the area of sensitivity theory, which addresses uncertainty in a system and its environment. “As a measure of feedback to cope with these uncertainties, we introduced a mathematical concept known as sensitivity,” said Perkins, who retired from the ECE faculty in 2005. “It measures the change in performance due to changing parameters or environment. Illinois became the center of this activity.”
At a conference in Dubrovnik, Yugoslavia, in 1964, Van Valkenburg met Petar Kokotovic, a Belgrade native who was completing doctoral work in Moscow. Kokotovic impressed Van Valkenburg with his command of the sensitivity literature from both sides of the Iron Curtain. Van Valkenburg arranged a U.S. tour for Kokotovic during the fall of 1965, which took the promising young researcher to the premier academic and industrial control research labs in the nation, introducing him to technical luminaries like Lotfi Zadeh, Richard Bellman, and Harold Chestnut.
“It seems that Mac Van Valkenburg was friends with just about everybody in the field, from the east coast to the west coast” said Kokotovic. “Mac launched me into a very high orbit!”
Later that year Kokotovic joined the Illinois control group, where he established himself as a leader in “singular perturbation” theory, which addresses the control of complex, dynamic systems whose behavior evolves in two or more time scales. Meanwhile, Cruz took the lead in applying theory of “differential games” to systems where multiple controllers, or players, seek to realize divergent goals. In the 1970s these two lines of research would merge into a Department of Energy–sponsored methodology for large-scale systems control.
Expanding the Lab’s Aim
The new control ideas percolating at Illinois had broad applications, and W. Dale Compton, CSL director from 1965 to 1970, spurred the group on to more interdisciplinary work. One such effort, led by urban and regional planning professor Jerrold Voss, drew upon control theory to achieve dynamic, interactive models of urban growth useful to planners. Other applications included economic policy and electric power systems. And after Compton left CSL to head the research division at Ford Motor Company, he contracted the Illinois group to deploy controls and electronics to improve automobile engine performance.
Beginning in 1981 with the hiring of Tamer Başar, the group’s ranks began to fill with a new generation of faculty who are today the senior members of the D&CL. Before coming to Illinois, Başar had already coauthored a book on non-cooperative game theory that has since become a classic in applied mathematics. At Illinois, he spearheaded research in the cutting-edge area of “H-infinity optimal control,” which addresses the old sensitivity issue from a new angle incorporating game theory. “You formulate it as a problem where the controller is playing a game against the uncertainty viewed as an adversary,” said Başar. “That way, you safeguard yourself against the worst that can happen.” In 1991 Başar coauthored a bestselling book on the subject, and today the book is still a standard text in the field.
In 1984 Kokotovic attended a conference in San Antonio at which a young mathematics professor from the University of Maryland–Baltimore County, P. R. Kumar, gave a talk addressing self-tuning regulators. The two struck up a negotiation, and by 1985 Kumar was a D&CL and ECE faculty member. He began to study manufacturing systems and developed scheduling policies for semiconductor fabrication that have been implemented at IBM’s Burlington, VT, plant. Later, Kumar turned his attention to wireless and mobile networks, developing some of the fundamental theory about the scaling of such networks. He and his students have written two of the most highly cited papers in that field.
The breadth of the control group continued to expand with the arrival of Sean Meyn in 1989. Meyn’s expertise, like Kumar’s, lay primarily in the area of adaptive control, the “optimistic” side of the control coin, in which the control law for a system can change as more is learned about the system function and its environment. Meyn’s work on nonlinear and stochastic systems led to a coauthored monograph that the current president of the American Economics Institute has called “the bible for economists who use Markov models.” Recently Meyn collaborated with U of I economics professor In-Koo Cho to improve the power market models and analysis used by economists.
At the same time, the early work on sensitivity has seen something of a renaissance lately in the field of robust control, which is the flipside of adaptive control in the sense that the engineer seeks to derive a control law that will function for an established range of system parameters. Perkins and Başar headed up the control side of a recent NASA-funded project to develop smart icing systems for small aircraft. “Icing changes the dynamics of the aircraft,” said Perkins. “We looked at sensitivity coefficients and monitored those online, and used those to decide when icing was occurring.” The project included faculty in aeronautical engineering and (because of the human factors involved in piloting) psychology.
The D&CL has supplied people as well as ideas to research institutions nationwide. Cruz retired from Illinois in 1986 and went on to become dean of engineering at Ohio State. After 25 years at Illinois, Kokotovic retired to become director of the Center for Control, Dynamical Systems and Computation at the University of California–Santa Barbara. In Fall 2008, robotics expert Mark Spong, a DCL member since 1984 and ECE faculty member since 2005, took the reigns as engineering dean at the University of Texas at Dallas. Among the many prominent control researchers who spent productive years with the Illinois group are Dmitri Bertsekas (1974–79) of MIT, Kameshwar Poola (1984–91) of Berkeley, and Jessy Grizzle (1985–87) of Michigan.
Decision and Control Today
In recent years, ECE faculty have continued to form a strong, though less dominant, contingent within an increasingly interdisciplinary D&CL. The newest ECE members of the lab are Christoforos Hadjicostis (1999), Daniel Liberzon (2000), Yi Ma (2000), and Seth Hutchinson (2007), an ECE faculty member since 1989 who adds to the group’s expertise in robotics.
Meanwhile, the non-ECE contingent within the D&CL has grown quickly. For many years it consisted solely of Spong and large-scale systems expert Juraj Medanic, both of whom were affiliated with the Department of General Engineering until its reorganization in 2005. Aeronautical engineering professor Petros Voulgaris joined the lab in 1993, followed in 1994 by mechanical and industrial engineering professor Andrew Alleyne. “Over the course of the 1990s it was great to see a real blossoming of interdisciplinary activity in the general area of decision and control,” said Alleyne of the diverse hires and increasingly broad-based collaborations since he came to Illinois. Today most of the lab’s faculty are non-ECE, a welcome development for all concerned.
Allerton Conference: A Lasting Legacy
In addition to pioneering research, the Decision and Control Lab also gave birth to the esteemed Allerton Conference on Communication, Control, and Computing, which originated in 1963 as the Allerton Conference on Circuit and System Theory. “Mac, Bill [Perkins], and I had conversations about the lack of conferences in our area,” recalled Cruz. “Mac said, ‘We will have our own conference at Allerton.’”
In the early days, Allerton was the only conference of its kind in the Midwest, offering researchers from across the country and abroad the opportunity to discuss the emerging control discipline.
These days, the annual gatherings have grown to encompass a broad range of engineering disciplines, including communication systems, queing networks, and VLSI architectures, just to name a few. Hundreds of top researchers from around the world come to enjoy—for a few a days in the beautiful setting of Allerton Park—the collegiality that Van Valkenburg established at CSL and that prevails in the lab to this day.