12.4 — CAD Software in Control Education and Research

12.4 — CAD Software in Control Education and Research

12.4 - CAD SOFTWARE IN CON TRO L EDU CAT ION AND RESEARCH Chainna n and Organiz er: Nairn A. Kheir The Univers ity of Alabama Huntsv ille, Alabama , ...

582KB Sizes 0 Downloads 35 Views

12.4 - CAD SOFTWARE IN CON TRO L EDU CAT ION AND RESEARCH Chainna n and Organiz er: Nairn A. Kheir The Univers ity of Alabama

Huntsv ille, Alabama , USA. Panelis ts : Karl J. Astrom Lund Institu te of Technol ogy Sweden. Gary B. Larront

Air Force Institu te of Technol ogy Dayton, Ohio, USA . Francoi s F. Cellier Univers ity of Arizona Tucson, Arizona , USA. C. Magnus Rimvall ETH/Ins ti tute of Automat ic Control Zurich, Switzer land.

Five formal present ations were made well discuss ion in a a lively se5sion (about 65 attende es).

by

Astrom gave a summary of Profess or K. Compute r-Aided Control Systems Design (CACSD) at the Lund Institu te of Technol ogy, Sweden, over the last 15 years. He emphasi zed the value of the extensi ve develop nent of CACSD tools in their graduat e instruc tional and researc h program s. Dr.Astro rn's group has succeed ed over the years in getting externa l funding for such develop nents. Present ly, these tools are used also on the lmdergr aduate level, where over 300 student s enjoy a combina tion of lecture s, CACSD exercis es, and lab experim ents. The present activit ies at Lund include develop nent of expert systems and users ' interfac es for modelin g and control ler design.

followe d attende d

Prof. Nairn Kheir highlig hted the motivat ion for organiz ing this panel discuss ion; to date there have been three symposi a on CAD of Control Systems sponsor ed by the IFAC in additio n to three more sponsor ed by the IEEE. The first IFAC symposiu m was held during 1979 in Zurich. The tremend ous interes t in this develop ing field since the early 1970's in univers ities (across Europe and the U. S. ) , as well as in industr y , softwar e of large number a in resulte d of package s(1-5) as well as a good number confere nces in present ations and session s Kheir Congres s . 1987 IFAC this includin g proceed ed to present a summary of the 1985 IFAC symposiu m(Lyngb y, Denmark ), as an illustra tion of such growth in researc h and develop nent, and sllllmal'i zed areas of conm::m concern in future develop nents . Such areas include : efficien cy of numeric al solutio ns, extenda bili ty , portabi li ty and user friendl iness. It is foreseen that more emphasi s on the use of CAD softwar e in the and in in researc h, classroo m enviroru nent, (expert knowled ge-base with connect ion systems ) (6-8) will continu e to grow especia lly that the computi ng power and cost of PC's and graphic al worksta tions continu e to favor such growth.

Cellier , of Uni versi ty of Profess or F . Arizona , Tucson, USA, discuss ed the importa nce of standar dization of interac tive CACSD package s. He pointed out that even the smalles t differen ces between the present defacto standar d, MATLABbased package s, makes an exchang e of data and conman d-like coded algorith ms very difficu lt . He suggest ed a standar dization on the level of operato rs(for enterin g and operati ng upon data), procedu ral languag es(for describ ing algorith ms), data and definit ions functio nal represe ntation (for exchang ing data) . Cellier stresse d the importa nce of making these standar ds open to extensio ns so as not to hamper future innovat ions and develop nents , and gave several example s on how such standar dization specifi cations would look like.

One can safely state that what is being wi tnessed in CAD of Control Systems represe nts just " a scratch on the surface "; however , one must keep an eye on the impact of such continu ing develop nent on the extent of depth and breadth of graduat ing future knowled ge in theoret ical control enginee rs and system theoris ts.

Profess or Larront ' s present ation focused on the role of the expert systems in CAD of Control Systems . Such expert systems model both the "design the as well as process " "design called collect ively are which knowled ge" intellig ence(fa cts, rules and process es).

107

108

N. A. Kh e ir

Dr. M. Rimvall of the Swiss Federal Institute of Technology(ETH), Switzerland, discussed the uge of CACSD tools in control education, and in particular the implementational requirements to be ~t on such packages. He made the distinction between CACSD in control education and education in CACSD . In the first case, CASCD exercises are included to illustrate theoretical lectures . In the latter case, numerical, algorithmical and implementational issues are of importance. Therefore students nllst be given access to well-structured modular packages to work with. Moreover, these classes may not focus mainly on progranming, and/ or software engineering topics. For reasons of modulari ty, data and program stucturabili ty, error-handling and portability, Dr. Rimvall favored the use of the progranming language Ada. QIA5tions

ana. Answers1from

Audience) :

Professor A.G.J . MacFarlane of Cambridge, UK, pointed out that a distinction has to be made between formalization and standardization . Expert systems and knowledge-based systems are essentially about formalization . What makes formalization important is that it implies sharability of knowledge, which is important in education. Therefore, one should ~t the highest priority in standardization on issues, which helps such sharability of knowledge, e.g . standardization on an appropriate functional language with a rigorously defined syntax . Dr . Samir Kotob of Kuwait observed that one of the main problems in CACSD systems is the nonstandardization of the user interface . Dr . Kotob presented the view that any standardization will not come from corrmi tt.ees but rather from a "Defacto Industry Standard ," and drew an analogy to the appearance of the "spread sheet" environments of the financial analysis field . Dr. Rimvall responded that such a defacto standard, in fact, already exists in CACSD(the different matrix-envirorunents based on MATLAB) but those slight differences in the uger interfaces of these packages make the exchange of data and conmand files difficult. Here a unifying effort is needed and would be welcomed by users of these packages. Dr. Cellier added that such a unification would be in the best interest of those producing CACSD software as well as the users. He also made a comparison with the (IBM) PC hardware and software market which prospers mainly because a standard exists . Dr. Kotob continued to ask the panel if we now witness the appearance of new workstations for CAD, and if particular graphical workstation is emerging as a standard. Dr . Astrom replied that, presently, there does not exist any standard graphical workstation envirorunent . Xwindows might be adopted as a de facto standard, but it is too early to make any defini te statement in this direction. The panel agreed that the field of CACSD, as we know it today , does not justify supporting the development of any dedicated, special ~se CACSD machine .

Dr . Hanselmann from Paderborn, West Germany, asked : How many CACSD groups are presently known to work with Ada? Dr. Rimvall replied that 4-6 groups are known to "seriously" use Ada for the implementation of interactive CACSD software packages (out of about 40 groups worldwide). Dr. Cellier POinted out that the number may actually be considerably larger as many projects are funded by the U.S .Department of Defence (which have to be implemented in Ada); the existence of these projects are not widely known .

Professor E. Ostertag from Strasbourg, France, drew a parallel between Ada and C, two languages both threatening FORTRAN as the most widely spread language for technical applications. He pointed out that C will also be standardized in the near future and asked if a wider accessibility of C might make C the more po~lar language of the two. Dr . Rimvall replied that the main disadvantage of both these languages was the lack of libraries with basic numerical algorithms. Presently, several groups are implementing such libraries in Ada, and this might give Ada lead time not easily recovered by

C. Dr . L . Motus o f Tallinn, U. S . S . R. , pointed out that control algori thrn design is but a small part of control systems design. Major problems arise in merging separate control algorithms into a functioning system when implementing the control system(distribution, on-line corrm.mication) . Dr. Motus asked if the panel knew of any research efforts where control algorithm design and control system implementation are incorporated. Dr. Larront and Dr. Rimvall replied that, systems which automaticallY produce real-time code describing a designed controller already exist commercially and have proven to be effective . However, they may require specific hardware configurations which presently limit their general use. Professor D.Frederick of Rensselaer Polytechnic Insti t.ute, Troy, USA, s tated that the pricing of software i s becoming an increasingly important issue . Most software packages are sold on a CPU-license basis, and this becomes very expens ive for universities with c lassrooms filled wi th workstations . The panel agreed that CACSD developers nllst be persuaded to consider refornulating their pricing policy in this respect.

Dr. G. Larront was given a question from the audience, namely i f there was not a risk that students using intelligent CACSD packages based on expert systems actually learn less than those students using "dumb" CACSD packages, as the expert system may result in that the students think less . Larront replied that, this is indeed an educational risk, but that a carefully designed CACSD package still forces the student to think, and only aids him/her in the process of learning. Moreover, Kheir and Larront proposed limi ting the use of CACSD packages to about 10% of the total available time with the rest of the time devoted to traditional, theoretical lecturing. Nairn A. Kheir M. Rimvall

CAD Software in Control REFERENCES 1 . Jamshidi M. and Herget C.J. "Computer Aided Control North-Holland Publishing.

(editors), (1985) Systems Engineering",

2.Frederick D.K., (1985) "Software Sunrnaries," in Computer-Aided CQnt..rQl. ~ Engineering (Jamshidi M. and Herget C. J . , edi tors) , North-Holland Publishing, pp. 349-384 . 3.Cellier F.F. and Rimvall C.M ., (1988) "Computer-Aided Control System Design : Techniques and Tools," in ~ Model irur and Computer Sjlllllation. (Nairn A. Kheir editor) Marcel Dekker Inc ., appendix A, to appear. 4.Rimvall C.M., (1986) "The Extended List of Control Software ELCS", Swiss Federal Insti tute of Technology, No. 2., Zurich . 5.Astrom K. J . , (1983) "Computer-Aided Mcdeling, Analysis and Design of Control Systems- A Perspective", IEEE Control Systems Magazine . 6 . Birdwell J . D. et al., ( 1985) "Expert Systems Techniques and Future Trends in ComputerBased Control System Analysis and Design Environment," Pr<:x:eedings of the 3rd IFAC Symposium on CADGE . 85 pp . 1-8, Copenhagen, Pergammon Press . 7.James J.R., Federick D.K . and Taylor J.H. ,(1987) "Use of Expert Systems Programming Techniques for the Design of Lead-Lag Compensators" , lEE Pr<:x:eedings, Vol. 134 , Pt.D., No.3, pp . 137-144. 8.Lam:mt G. B., (1987) "Expert System developnent for Computer-Aided Control System Development" , Proceedings of the 10th IFAC Congress, Vol.l1, Munich .

AC-G

109