F E A T U R EF E A T U R E

A Student Control CompetitionThrough a Remote Robotics Lab

utomatic controlis one of the

t e c h n i c a ldomains in

which newtechnolo-

gies and tools for distancelearning have often beenapplied [1]. The AutomaticControl Telelab (ACT) is aremote laboratory for auto-matic control developed atthe University of Siena. Thisarticle describes a new ACTmodule referred to as thestudent competition system.

ACT has been used in undergraduate control systemscourses since 1999 [2]. The aim of the project is to allow stu-dents to apply their knowledge of control theory in an easyway, without restrictions due tolaboratory operating times andprocess availability. Currently,ACT is accessible 24 hours a dayfrom any computer connected tothe Internet. Key features of ACT are its simple user-friendlyinterface and the possibility of integrating user-defined con-trollers in the control loop of the remote process through a

MATLAB/Simulink model.ACT is continually upgradedwith new software versionsand experiments. In the pre-sent ACT release, fiveprocesses are available foronline experiments: a dcmotor, a tank, a magnetic lev-itation system, a two-degree-of-freedom helicopter, and aLego Mindstorms mobilerobot [3].

A new component of theACT experience is a studentcompetition, in which stu-dents or groups of students

compete to develop the best controller for a given remoteprocess. A typical competition session starts with definingcontrol system requirements for one of the ACT processes.

Students then access ACT anddesign controllers to steer theprocess during the competition.The ACT server stores the stu-dents’ controllers and data in a

database, computes the performance measures, and assigns ascore to the controllers. A ranking is computed and reportedto students as feedback for their learning process.

A new tool for demonstrating control design issues

By Marco Casini, DomenicoPrattichizzo, and Antonio Vicino

© PHOTODISC AND ARTVILLE, LLC.

February 2005560272-1708/05/$20.00©2005IEEE

IEEE Control Systems Magazine

A

February 2005 57IEEE Control Systems Magazine

Student Competition OverviewMost remote labs allow users to run experiments using pre-defined or user-defined controllers. Students can run anexperiment and look at the dynamic response, but informa-tion on the controller’s performance is not generally provid-ed, and it is not possible to know how controllers designedby other people behave on the same process. This deficien-cy is one of our main motivations fordesigning a student competition forthe remote laboratory. Through thistool, students are aware of the perfor-mance requirements that their con-trollers must satisfy. Moreover, aranking of the best controllers, as wellas the time plots of related experi-ments, are provided.

Features of the ACT competitionsystem are described in the follow-ing sections.

Remote ExercisesIn addition to controller synthesis exercises, this toolchallenges students to design a controller that must satis-fy performance requirements and allows them to testtheir controller on a remote real process. At the end ofthe experiment, the performance measures are automati-cally computed and shown to the user; if such measuresfulfill the requirements, the exercise is completed. Anoverall measure, usually a weighted sum of the previousmeasures, is then computed, and the controller is includ-ed in the ranking list.

Controller ComparisonThe competition ranking is provided to the students. Dur-ing the final competition lesson, students who havedesigned the best controllers are invited to discuss theirprojects, while the instructor explains why some controlstrategies work better than others.

Many Benchmarks on the Same ProcessIt is possible to provide more than one competition bench-mark on the same process, thus increasing the number ofremote exercises available for students. New benchmarkscan easily be added for implementation owing to the soft-ware architecture of the ACT competition system.

A Competition Session DescriptionIn this section, we describe a competition session involv-ing the magnetic levitation process. The requirements fora step reference input are as follows:

● the 5% settling time must be less than 1 second● the time response overshoot must be less than 10%.

To compete, students register by completing a form andsubsequently receive a username and password. The suc-

cessive steps during a competition session are summa-rized as follows:

● study the physical model of the process and lin-earize it around an operating point

● analyze the linear model through frequency domainplots and root locus techniques to determine appro-priate bounds for the controller gains

● synthesize a controller that achieves the given per-formance specifications by using lead-lag or propor-tional-integral differential (PID) controllers

● construct a Simulink model of the controller● simulate the controller● test the controller on the actual remote process● analyze the system response and, if not satisfactory,

adjust the controller and try again.To design the controller, students must download a

Simulink template file and connect the signals describingthe reference, output, and command signals with suitableSimulink blocks. Since the output signal is the sensorreading, the students must insert a block that models thesensor characteristic.

A special graphical interface allows students to describethe structure of their controllers, such as PID, and set the sam-ple time of the experiment. In addition, the user must specifythe file containing the controller and, if needed, the MATLABworkspace file (.mat) containing the data for his/her con-troller. These files are uploaded to the server, compiled, and, ifno error occurs, executed on the real remote process. A sec-ond graphical interface (Figure 1) allows the user to start theexperiment and observe its behavior through plots and livevideo windows. At the end of the experiment, the performancemeasures are automatically computed and displayed.

Since different controllers can satisfy the performancespecifications, an overall score is computed to generate aranking (Figure 2). This overall score is obtained as aweighted sum of the performance measures. If a controllerdoes not satisfy the performance specifications, an overallscore is not computed.

It is possible for each user to view a controller reportwith information on ranking and other data, such as thecontroller description, the nickname of the user, and theuser’s institution.

Control systems classes at theUniversity of Siena are using the

ACT competition system with greatinterest and excitement.

February 200558 IEEE Control Systems Magazine

Teaching ExperiencesIn this section, we describe the role of the student com-petition in undergraduate control system classes at theUniversity of Siena. The lecturer begins by illustrating the

physical model of the magnetic levitation system, empha-sizing its unstable and nonlinear dynamics, and suggest-ing how the students might linearize the dynamics todesign the controller.

Since ACT is accessible at all times, unless beingused by someone else, students are able to analyze thesystem, design candidate controllers, and test themprior to the second competition class. During the sec-ond class, the lecturer answers student questions andhelps solve typical problems. For example, the lecturermight suggest the use of a prefilter on the referenceinput to obtain smoother command signals and, as aresult, better performance.

At the end of the competition, most of the students areable to design a satisfactory controller. Student feedbackon the project has been enthusiastic, especially since stu-dents are able to apply many theoretical notions. More-

over, the students try to do theirbest to obtain a good ranking.However, the real motivation forthe competition is not merely todetermine a winner but rather togive students a new tool that canhelp them better understand prac-tical control design issues, whileincreasing their interest in controlsystems and technology.

ConclusionsIn this article, the student competi-tion of ACT is described. Controlsystems classes at the Universityof Siena are using the ACT competi-

tion system with great interest and excitement. This toolstimulates students’ interest in control systems andenhances learning by facilitating the understanding ofpractical control design issues. We are currently designingan ACT competition for a Lego mobile robot, which can beused by students of robotics courses. The ACT home pageis http://act.dii.unisi.it.

References[1] S.E. Poindexter and B.S. Heck, “Using the Web in your courses:What can you do? What should you do?,” IEEE Contr. Syst. Mag., vol. 19,no. 1, pp. 83–92, 1999.

[2] M. Casini, D. Prattichizzo, and A. Vicino, “The Automatic ControlTelelab: A user-friendly interface for distance learning,” IEEE Trans.Educ., vol. 46, no. 2, pp. 252–257, 2003.

[3] M. Casini, D. Prattichizzo, and A. Vicino, “The Automatic ControlTelelab. A Web-based technology for distance learning,” IEEE Contr.Syst. Mag., vol. 24, no. 3, pp. 36–44, 2004.

Key features of ACT are its simpleuser-friendly interface and thepossibility of integrating user-definedcontrollers in the control loop of theremote process through aMATLAB/Simulink model.

Figure 2. Controller ranking. At the end of the experiment, per-formance measures are automatically computed, and the con-troller is included in the ranking list.

Figure 1. The interface showing the running experiment.Through this interface, it is possible to start and stop theexperiment, as well as view data plots and online video.

February 2005 59IEEE Control Systems Magazine

Marco Casini (casini@ing.unisi.it) received the Laureadegree in computer science engineering and the Ph.D. incontrol systems from the University of Siena in 1999 and2003, respectively. He is currently a research associate atthe Dipartimento di Ingegneria dell’Informazione of theUniversity of Siena. Since 1999, he has held several fellow-ships at the Dipartimento di Ingegneria dell’Informazioneof the University of Siena. In 2001, he was a visiting scien-tist at the Laboratory for Information and Decision Sys-tems (LIDS) at the Massachusetts Institute of Technology,Cambridge. His research interests include distance learn-ing, remote laboratories, set membership identification,and identification for control. He can be contacted at theDipartimento di Ingegneria dell’Informazione, Universitàdegli Studi di Siena, Via Roma 56, 53100 Siena, Italy.

Domenico Prattichizzo received the M.S. degree in elec-tronics engineering and the Ph.D. degree in robotics andautomation from the University of Pisa in 1991 and 1995,respectively. In 1994, he was visiting scientist at the Arti-ficial Intelligence Lab at the Massachusetts Institute ofTechnology. Since 2002, he has been an associate profes-sor at the University of Siena. He was cochair of the Sec-ond IEEE International Workshop on Control Problems inRobotics and Automation in 2002. His research activitiesinclude distance learning, optimal control, robotic manip-ulation, and visual servoing. He is an associate editor of

IEEE Transactions on Robotics and IEEE Transactions onControl System Technology, an editorial board member ofthe Journal of Dynamics of Continuous, Discrete and Impul-sive Systems Series B, and a member of the ConferenceEditorial Board of the IEEE Control Systems Society. He iscoauthor of more than 100 papers in the area of controlsystems and robotics.

Antonio Vicino received the Laurea in electrical engi-neering from the Politecnico di Torino, Torino, Italy. From1987 to 1990, he was associate professor of control sys-tems at the Università di Firenze. In 1990, he joined theDipartimento di Ingegneria Elettrica, Università diL’Aquila, as professor of control systems. Since 1993, hehas been with the Università di Siena, where he is current-ly dean of the engineering faculty. In 2000, he founded theCenter for Complex Systems Studies of the University ofSiena, where he is currently director. He was an associateeditor for IEEE Transactions on Automatic Control from1992 to 1996. Presently, he is an associate editor for Auto-matica and associate editor at large for IEEE Transactionson Automatic Control. He is a Fellow of the IEEE and authorof 190 technical publications. He is also coeditor of twobooks on robustness in identification and control. Hismain research interests are in robust control of uncertainsystems, robust identification and filtering, applied sys-tem modeling, and mobile robotics.

Job AdvancementA classic example is said to be that of Humphrey Potter, a twelve-year-old boyemployed in the early eighteenth century in England to operate a handle thatadmitted steam to the cylinder of Newcomen’s first steam engine. Potter, being aperson of considerable intelligence, yet having nothing else to do but open and closesteam valves, noted that whenever the piston was at one end of the cylinder heopened one valve, and whenever it was at the other end he opened another to let thesteam into the other side of the piston. He observed the piston-valve relationship,hooked the valve to the piston so that the piston operated the valve automaticallyand thereby invented the slide valve mechanism which is still in use today. Inciden-tally he worked himself out of a job and into a better one.

— Quoted from W. Buckingham, Automation, Its Impact on Business and People, Harper, 1961. © EYEWIRE