»
P U B L I C AT I O N A C T I V I T I E S
Some Thoughts About Publishing Results in Our Field
T
his issue of IEEE Control Systems Magazine contains some comments from the editor-in-chief of IEEE Transactions on Automatic Control (TAC), Panos Antsaklis. These comments have been adapted from his October 2013 editorial in TAC and are of broader interest than to just the readership of TAC.
T
here are great practical needs in control that are being addressed but unfortunately not necessarily by using the latest theoretical developments. The reason is that many papers reporting theoretical results are written to be appreciated primarily by the experts in that particular area, and they are typically difficult to understand and inaccessible to other readers and practitioners. I understand that it is not easy to write articles that meet the high academic standards necessary for acceptance in IEEE Transactions on Automatic Control (TAC) while at the same time being written to be accessible by a wider readership and accomplishing both requirements in the few number of pages allowed. Having said that, I believe we can do a better job of making our results more accessible and in convincing people of their importance and encouraging them to use them. I have tried to encourage authors to do exactly that by requiring that their contributions be put in context so as to be better understood and appreciated. For this, authors of TAC are asked to include a paragraph in the introduction describing clearly and in some detail the connection of the topic dealt within the paper to systems and con-
Digital Object Identifier 10.1109/MCS.2013.2279448 Date of publication: 14 November 2013
trol problems. If the study was motivated by an application, this should be mentioned. If the motivation is to extend previous mathematical results, this fact needs to be stated. In that context, it should be explained why the results are of importance. It is much easier for the reader who is not an expert in that particular area to understand and appreciate the importance and usefulness of the results by connecting to a class of applications. This is most important because our field is very broad, both in the types of problems addressed and the approaches used. The mathematical tools used vary widely, from ordinary differential equations (ODEs) and partial differential equations (PDEs), to automata and Petri nets, to logics, to optimization algorithms in continuous- and discrete-time settings. So if the reader does not completely understand the mathematics, we have to make an extra effort to explain the significance and implications of our theoretical contributions. One way to encourage others to use our results, whenever possible, is to go beyond our theoretical contributions and start addressing the next step, that of implementation. We should, whenever possible, suggest algorithms and offer software code so potential users can easily try out the results. I assume that we all want our results to be used. I do not assume that all of our results
22 IEEE CONTROL SYSTEMS MAGAZINE » DECEMBER 2013
offer the opportunity to give rise to algorithms, but some certainly can and should. And now here is a more difficult task. We need to address bigger problems. A system typically is more than a set of ODEs, and the specs may not be conveniently described in the frequency domain. While these specs served us well in the past and are still very useful, we need to move on because the problems and their descriptions have become much more sophisticated and much more demanding. We are in the systems area after all, an area that prides itself for considering a wider view of the problem, taking a system’s or bird’s eye point of view. In hybrid dynamical systems we combine discrete and continuous dynamics to study system behavior. Today we have data and lots of it that need to be considered together with our mathematical models. What is the best way to go? This is quite a challenge. In my current research, I study cyberphysical systems (CPSs) that combine cyber, physical dynamics, networks, and perhaps human operators. Because of demanding specifications, tight interactions between cyber and physical dynamics need to be designed and explored. We typically do not have exact models because the CPS may expand and contract during its life cycle. In my group, we have been using the energy-like concepts of passivity and dissipativity in new ways to characterize CPSs, and we hope to derive a methodology to design safe CPSs. So here is a very general framework where we use extensions of rather classical control concepts to address (continued on p. 41)
conditions is an interesting open area for control research. Over the years, many control methods have been developed that have certain robustness properties for systems within some range of uncertainties but not specifically for abnormal conditions. The uncertainties of control systems under abnormal conditions may be too much for regular control designs to handle. Many such abnormal conditions are caused by system faults, such as actuator and sensor failures, and structural damage. Recently, new research efforts have been made in our community to develop resilient control system technologies for performance-critical and emerging control system applications, in which uncertain system faults are of main concern. A resilient control system is required to have the capacity to maintain certain desired performance in the presence of uncertain multiple faults. New control theory and techniques are needed to guarantee desired system resilience, that is, to increase the system’s ability to recover from uncertainties and faults. I believe such research has theoretical importance and practical urgency; in particular, there are new control theoretical problems to be solved.
»
P U B L I C AT I O N A C T I V I T I E S
the problems. We also have developed ways to experimentally determine passivity indices; we have also worked on algorithms and written software so interested parties can easily and conveniently try out our results.
Some Closing Thoughts At our TAC editorial office, we receive a steady stream of excellent papers in the very broad area of systems and control, and we apply high reviewing stan-
Profile of Gang Tao • Current position: professor of electrical engineering, University of Virginia. • Visiting position: visiting professor at Nanjing University of Aeronautics and Astronautics, China. • Contact information: Department of Electrical and Computer Engineering, University of Virginia, Charlottesville, Virginia 22904-4743 USA, +1 434 924 4586,
[email protected], http://www.people.virginia.edu/~gt9s. • IEEE Control Systems Society experience highlight: associate editor, IEEE Transactions on Automatic Control, 1996–1999. • Notable award: IEEE Fellow, 2007.
Q. You are the author of four books in the control field. What topics do these books cover? Gang: The books are on adaptive control, covering basic theory and some advanced research topics. My 1996 Wiley book (coauthored with P.V. Kokotovic) is on adaptive control of systems with actuator and sensor nonlinearities. My 2003 Wiley book is a textbook on adaptive control theory for both continuous-time and discrete-time systems, written largely for multivariable systems. My 2003 Springer book (coauthored with A. Taware) is on adaptive control of systems with sandwiched nonlinearities, and the topic of my 2004 Springer book (coauthored with S. Chen, X.
Tang, and S. M. Joshi) is on adaptive control of systems with uncertain actuator failures. I also coedited a book on adaptive control of systems with nonsmooth nonlinearities (with F.L. Lewis in 2001; Springer) and a book on advances in control systems (with J. Sun in 2009; USTC Press).
Q. What are some of your interests and activities outside of your professional career? Gang: I like to read history books and articles to learn the details of some historic events and people of interest.
Q. Thank you for your comments. Gang: I appreciate this opportunity. Thank you!
(continued from p. 22)
dards to make them even better. We are doing well; we publish a large number of excellent papers containing significant results. But perhaps we should also gaze down the road to see what happens next. Perhaps we do not go far enough in our pursuit of solutions. Maybe we stop too soon. Or maybe we are not thinking big enough. Are we letting others think for us and take ownership of problems? Are we leaders in technology? Can we perhaps do better? Our formulations
do not address the whole problem. We address parts of the problem, typically the part that is described by ODEs. And in the part we are addressing, we typically stop short from designing algorithms, let alone implementing algorithms in software. So this is something to think about if we want to have a vibrant, relevant, and influential research community. Panos Antsaklis
DECEMBER 2013 « IEEE CONTROL SYSTEMS MAGAZINE 41
