Enhancing Learning by Integrating Theory and Practice | MathWorks Research Summit

Thank you also for this opportunity to show some recent tools we are using for an integration of the teaching in control engineering courses. So I will describe very quickly the structure of our engineering courses, in particular focusing on automatic controllers is the topics I'm teaching. And then some simple example of how our teaching methods. In particular, the main problem is how to engage students.

So I will use some methods we try to do it. And finally, we will verify the sources of our methods using well-established performance matrices. This is our main building, which is a sugar factory which was transformed into our department. And up to 2006 we had one Bachelor program which consists of, in the first two years, we had two mainly basic and only at the second year we started with the curricula courses.

And only at the third year we have, again, two curricula courses. And at the master program we have only two curricula courses. But since the last year-- from the last year, we modified the structure of our program and now we have a Bachelor program, which in the first two years, we have-- the first year with the basic courses.

In the second year we have an option between two courses. And now at the third year the students have to choose among four courses, so electronic engineering, computer science, telecommunication, and automatic control. And then at the master program we have four curricula each, so it's really challenging for trying to organize at our best structure.

So by focusing on the control topics, I'm teaching also with the other two colleagues of mine. The second year we have the basic of automatic control and the third year we have the digital control, and teaching this digital control with other system identification, electric drives, and industrial automation. At the master we are advanced control techniques like multivariable control techniques, supervision and adaptive systems, which consists of also the use of neural network fuzzy logic.

So we had to really seriously think about the structure and how to optimize our lectures. So now we have the problem of how to integrate the theory and practice for enhancing the learning to our students. So of course, obviously it's a difficult task because every year you have different quality of students like the wine.

So depends also the quality of the previous courses, the lecture distribution, because every year the lectures can be moved from one period to another one, and also the number of students. So it should be really an adaptive system, and every year you have to change or maybe to modify your teaching methodology. So how to engage students, which is the main problem.

Of course, you can-- there are the easiest options are to use dirty jokes so you can attract the attention of students. Or the other option is because we are in Italy, just talk about the football teams and everybody is really excited. But of course, it's too easy. So we use the first three teaching strategies. The first one is quite simple. So it means that they have to learn by doing.

So most of my lectures are in the laboratory working with the computer, and we try to organize also the lectures using manual and semi-automated design procedures so they can drive the students to understand also the theory and the practice. And another important point is to use realistic and the real examples because the main problem with students, if you talk to them about some general things they are not really interested.

So it's just talk to them about concrete things which is a very important point. And the tools now. For the first year we move Matlab at the second year of our bachelor's degree and the Simulink at the third year. And I will show some examples of learning and teaching tools we are using, for example, in automatic control and digital control. So let's see some example.

The first one, which is the one we introduced for the first year, is the design of a lead or a lag network. So a very established approach, it was using an empirical design procedure because it means that you have to select the crossover frequency where you want to modify the phase margin of the control system and then you have to play with the lag, for example. I put the example of the lag network trying to move in the optimal way the frequencies of the pull and the zero.

Of course, for the students it was quite hard to understand where to put in order to avoid, for example, the side effects of the network were not at the best position in both the diagram. For the first time we introduced the inversion formulas that can be more complicated, but by integrating them in a graphical interface, they help the students to understand the range where they have to select the optimal position of the lag network.

So in this way they have also the feasibility region where the network has to be designed and they very simply obtain it by solving this equation I put there to understand, which represent also the fact that in this range of frequencies you can perform the compensation of the phase margin you want to design. And then the students can use the bold diagram for verifying their design.

And the second example I bring it to you is how to derive the transfer function from a simple circuit. Of course, take into account that we are teaching to both students from information science and also from mechanical engineer, so you have to balance between, because of course, the mechanical engineer students can be bored by the use of all the pictures of electrical circuits.

So we use the introduce the symbolic toolbox because it's very also an easy way for deriving just by representing the equations, the differential equations as a symbolic function. So you are able to solve this equation and to derive in an automatic way the state space and then the transfer functions of form and by substituting in an automatic way the parameters of the system. So it's very easy for the students to understand.

And of course, in order to also present some interesting examples for the mechanical engineer students, we took-- for example, we proposed to derive the model of the joints of these lightweight rubber too. It comes from the Institute at DLR. And again, using the differential equations is very easy to derive the state space formulation of this model, of course for the controller design.

Another way of learning by doing for students is, for example, when digital control is important that they can understand the difference equations, which is maybe, from a theoretical point of view, it's quite hard to understand what is the difference equation. So they need to know what is it because it is what they are going to implement when they want to realize a real control law in microprocessor.

So I use this, for this example, for introducing the concept of vector stability of discrete time systems. Also the notion of the Matlab arrays, graphs, and also the Simulink blocks. And moving to the design of controllers, for example, the PID controller-- so I teach the students to use so they can revise also the concept of the route locks for driving the critical gain of the system for designing the PID, for estimating the PID parameters, for using the Ziegler-N equals and the oscillation method.

So it's very easy. And also compare with the autotuning procedure proposed by Matlab already implemented in the Simulink block of the PID. So they can compare the performance obtain it with the Ziegler-Nichols auto tuning procedure and the autotuning procedure proposed in Matlab, and also to understand advantages and drawbacks of the two methodologies.

And the other examples we are proposing, this has come from a cooperation with the local companies. Now it is owned by FCA, and previously it was Fiat and then is owned by FCA. And so we have real data acquired from a real Jeep Wrangler engine. So they can use the real data, for example, practicing using neural networks, fuzzy logic, the ANFIS tool, and the neural network tool, which is very important tools we are using.

And again, note other realistic applications we are deriving from European projects. So for example, this is the simulator developed in the Simulink environment of the Mars Express. It was used for showing the design of an observer. So let's see. Let's try to conclude these recipes I'm proposing. So I suggested to do learning by doing realistic and real examples and some design procedure to drive the students.

Another fourth point which is quite important to me is the teamwork, but not really the teamwork used also in the assessment and in the project-- the final project of the students. But in the sense I will show you in two slides. The first one is the teamwork assignment. The problem when you proposed you organize the students in teams is the problem is how to select the composition the optimal composition of the team.

So maybe Frank is going to talk about it in his talk. But in my case the teamwork is used for peer review. What does it mean? It's a kind of active learning we were discussing in the working groups this morning and also yesterday. I mean that I propose the students some exercises and they have to solve the exercise, but at the same time they are selected randomly they have also revised the exercise done by the other one.

So in this way they have to be prepared to evaluate the exercise made by the other students. So in this way, in my view, it's a way of an automatic-- of an active learning approach to this kind of evaluation. So the last part, this last slide are devoted to the verification of our approaches. So we have a national committee which is evaluating the quality of our courses.

The university in Italy has to be to receive the approval by the ministry. So there is a committee which is called the AVA, which means self evaluation and evaluation and approval. So every university, if every three years you don't have this kind of approval, you cannot open. You cannot teach, so you cannot deliver your degrees.

And so therefore this quality assessment, the students before the exams, they have to fill some questionnaire, like the questionnaire you have to fill before leaving this event. And they are based on the number of 12 questions. Every questions there are two or three questions regarding the teacher quality, the course structure, the course contents, and also the quality of the hands on and the laboratories.

And every question has a score between 0 and 10, and all the results are collected into a national database. So for each course every university has the result. This, for example, represent-- the circle represents the 12 questions and the score for each question. So a very good course, for example, at the left-- in the right, it has almost all the points near to the center because it means that the students is an average of all the answers of the students.

On the left, you have the course we were teaching in 2006 before introducing this new methodology. So these results show the effectiveness of our tools, which we are using in our courses. So this is a validation. When you want to validate a controller, this is a validation of our tools. This is an example of one of our courses. So for concluding, I tried to show you to describe some learning and teaching methodology we are proposed to our students that are used for both the learning, teaching, and also the self evaluation. And also thank you for your attention.

Thank you.

[APPLAUSE]