The Community for Technology Leaders
Green Image
Intelligent control
Control is the act of affecting a dynamic system to accomplish a desired behavior. In developing control design theory over recent decades, control engineers have used well-defined system goals to produce many effective techniques for implementing a wide range of system controllers. The newly coined term intelligent control refers both to the control design approach or philosophy and to implementation techniques that emulate certain characteristics of intelligent biological systems.
Intelligent control as a discipline owes much to both computer science and computer technology. For instance, engineers use software development and validation tools for expert systems to construct expert controllers that automate the actions of a human operator controlling a process. Other knowledge-based systems such as fuzzy systems (rule-based systems that use fuzzy logic for knowledge representation and inference) and planning systems (which emulate human planning activities) find similar use in automating the perceptual, cognitive (deductive and inductive), and action-taking characteristics of humans who perform industrial control tasks.
Artificial neural networks crudely emulate biological neural networks. For instance, they have been used to learn how to control systems by observing the way that a human performs a control task. Alternatively, neural networks can learn in an on-line fashion how to best control a system by taking control actions, rating the quality of the responses achieved when these actions are used, and then improving system response by adjusting the recipe used for generating control actions.
Developers use genetic algorithms to evolve controllers via off-line computer-aided design of control systems or on-line by maintaining a population of controllers and using survival-of-the-fittest principles— fittest being defined by the quality of the controller's response. In addition to software development tools, several computer chips are available for fuzzy, genetic, and neural systems that provide efficient parallel implementation of complex reasoning processes in real time.
All these approaches fall into the realm of intelligent control. In each, both the biological background and the effect of computer science and computing technology drive developments of the field of control by providing alternative strategies for the functionality and implementation of controllers for dynamical systems. In fact, the control field is moving toward integrating the functions of intelligent systems, such as those just listed, with conventional control systems to form highly autonomous systems that can independently perform complex control tasks—such as for robotics and vehicular applications. This trend toward intelligent autonomous control systems is gaining momentum as control engineers solve many existing problems and naturally seek new control problems that involve broader issues and require the full capabilities of available computing technologies.
The development of such sophisticated controllers does, however, still fit within the conventional engineering methodology for constructing control systems. This development involves mathematical modeling using basic principles or data from the system, along with heuristics. Some intelligent control strategies rely more on the use of heuristics (such as direct fuzzy and expert control). Others use mathematical models in the same way they are used in conventional control. Still others use a combination of mathematical models and heuristics.
Needed now are systematic controller construction methodologies that apply to a wide class of applications. Some methodologies for constructing intelligent controllers are quite ad hoc—such as for the fuzzy or expert controller. Still, they are often effective because they provide a method and formalism for incorporating and representing the nonlinearities needed for high-performance control. Other methodologies—such as for certain neural and adaptive fuzzy controllers—are no more ad hoc than ones for conventional control.
To help verify the performance of the control system being implemented, control engineers still need analysis of stability, controllability, and observability properties. Although we have seen significant progress recently in stability analysis of fuzzy, neural, and expert control systems, much more work is needed in nonlinear analysis of intelligent control systems. Simulations and experimental evaluations of intelligent control systems are needed to validate performance levels. Also needed are engineering cost-benefit analyses involving the evaluation of competing control techniques, performance, stability, ease-of-design, lead-time to implementation, complexity of implementation, cost, and other issues. Ultimately, the careful evaluation of intelligent control approaches will help attenuate the hype that sometimes surrounds them and lead to the best-possible engineering solution.
Although the intelligent control paradigm typically focuses on biologically-motivated approaches, it is not clear that the resulting controllers behave drastically differently—they are not mystical; they are simply nonlinear, often adaptive controllers. This is, however, not surprising, because every new intelligent control approach seems to have a precursor in an existing conventional control approach (see Table 1). While some new concepts grow from the field of intelligent control, a crucial role certainly remains in evaluating and developing the emerging field of intelligent control for the control engineer and control scientist with the standard control background.
Regardless of the technique or methodology used, the ultimate proof lies in the method's success in actual applications. While still relatively young, intelligent control methods have found various applications in robotics, manufacturing, automotive systems, underwater vehicles, ships, spacecraft, aircraft, process control, engine control, missiles, weapon control, automated highway systems, and elsewhere. All these areas, of course, have reliably used conventional control techniques for many years.
The important question is then, what, if any, will be the niche of intelligent control? What applications is it good for, and not good for? Will intelligent control simply provide some techniques to add to the standard control engineering toolbox? Or will intelligent control provide a whole new philosophy for approaching the design and control of large complex dynamical systems? This Special Track should help answer these questions.
The Special Track on Intelligent Control
As the variety of topics we have just discussed demonstrates, intelligent control is quite diverse and wide-ranging. Traditionally, the fields of control, artificial intelligence, and operations research have made major contributions. Many other fields, cutting across a variety of disciplines, are also involved—electrical engineering, computer science, mechanical engineering, chemical engineering, aeronautical engineering, industrial engineering, mathematics, psychology, and so on. The wide variety of relevant topics and viewpoints makes it somewhat difficult even to define the boundaries of the field and agree on the essential ingredients of what we mean by intelligent control. (For some recent progress in this direction, see our list of recommended readings.)
In this Special Track we seek to
  • promote interactions between all the disciplines interested in intelligent control,
  • provide access to some of the latest promising research in a wide spectrum of the areas of intelligent control, and
  • open a window on the techniques of intelligent control for the practicing engineer who is interested in pursuing the ideas further.
The Special Track on Intelligent Control appears in this and the next issue of the IEEE Expert.

Table 1. Related conventional and intelligent control techniques.


"Defining Intelligent Control," Report of the IEEE Control Systems Society Task Force on Intelligent Control (P.J. Antsaklis, Chair), IEEE Control Systems Magazine, Vol. 14, No. 3, June 1994, pp. 4-5, 58-66.

Intelligent Control: Theory and Practice, M.M. Gupta and N.K. Sinha, eds., IEEE Press, Piscataway, N.J., 1995.

Handbook of Intelligent Control: Neural, Fuzzy, and Adaptive Approaches, D.A. White and D.A. Sofge, eds., Van Nostrand Reinhold, New York, 1992.

An Introduction to Intelligent and Autonomous Control, P.J. Antsaklis and K.M. Passino, eds., Kluwer Academic Publishers, Norwell, Mass., 1993.

Kevin M. Passino is an associate professor in the Department of Electrical Engineering at Ohio State University. His research interests include intelligent and autonomous control techniques, nonlinear analysis of intelligent control systems, failure detection and identification systems, and scheduling and stability analysis of flexible manufacturing systems. He received a PhD in electrical engineering from the University of Notre Dame in 1989. He is an associate editor for the IEEE Transactions on Automatic Control. He is a member of the IEEE Control Systems Society Board of Governors; the IEEE Computer Society; the IEEE Systems, Man, and Cybernetics Society; and the IEEE Society for Social Implications of Technology. Reach him at the Dept. of Electrical Engineering, Ohio State Univ., 2015 Neil Ave., Columbus, OH 43210-1272;;
Ümit Özgüner is a professor of electrical engineering and director of the Center for Intelligent Transportation Research at Ohio State University. His areas of research interest are in decentralized and hierarchical control for large-scale systems and mechatronic systems in general, and applied automotive control, flexible structure control, and automated highway systems. He received his PhD from the University of Illinois in 1975. He is Technical Committee Chair on Intelligent Control for the IEEE Control Society, represents the Control Society in the IEEE TAB Intelligent Transportation Systems Committee, and is chairing the Ohio Aerospace Institute Technet on Control. Reach him at the Dept. of Electrical Engineering, Ohio State Univ., 2015 Neil Ave., Columbus, OH 43210-1272;;
335 ms
(Ver 3.3 (11022016))