 So basically it's partitioned in two parts. Very quickly I'll run through the academic part and then look at how the research applications of it is there. So basically we use it in our academics in the area of systems modeling. Most often we tend to use analytical methods by which is the best way to start teaching mechanics and system dynamics modeling and ask students to write codes themselves and learn from it. There is a tendency now for students to use, if you want to do quick problem solving of advanced applications in, directly in the application scenario, then we use tools like MATLAB and SyLab has been used sparingly, I would say. We also use one of the symbolic modeling and simulation language based software, which is based on bond graphs. I don't know how many of you would have heard of it, but it is a tool developed to model physical systems. And that's more a symbolic modeling tool. You can put in the Google reference to it and get sites. But I will not talk about it in this workshop here, because that's more a symbolic thing and it does not come in the purview of our numerical simulation methods. So if you're talking of teaching people, we are trying to bring in mechanical engineering, education, trying to make people be able to understand dynamics of physical systems and see how they can simulate various kinds of systems. So you may be looking at some very simple systems like this, wherein if you have a constrained mechanical systems where there's motion and linkages, you have a variety of elements. The best way is to use analytical methods to derive the equations, the Lagrangian methods, and come up with the system equations and then solve it by any means. First principles we focus on trying to get the equations right, and then we try to see the solutions for this. MATLAB, SILAB, and all these other tools right now for that. We also look at courses like dynamics of processes, controls, applications, wherein it has a component on the systems modeling, which primarily uses the state space approach, and that's very easily handleable by the softwares, both like MATLAB and SILAB. We have seen that this morning a few examples, and I'm sure in the afternoon we'll have more of it. So I don't want to go through this in details. But as we go towards the controls part, we look at more aspects like controllability, observatory analysis, control systems design, designing of the controllers, tuning of the controllers, et cetera. And I specifically more target these towards applications in manufacturing and robotics in our courses. So basically they work with systems directly in the physical domain when we are teaching like this, like work with mechanical systems or electrical systems right from there, just when we try to implement them in psych for their simulation purposes. I'm glad to hear today morning that we can actually have an interface to psych us with the modelica and actually do this online. Otherwise, our alternative is we use bond graphs for it as well. So when we design control systems, we actually expect students to learn through various aspects of modeling, controlling the loop, and look at its response with various parameters. So I'll quickly now jump into the research applications. I'll pick up three applications which I found it very useful, our students do use it for these kind of tools. And I think Sylab will be very useful here too. So quickly I'll go through this. The first one on sensor network applications wherein the objects of our work, research work is to track objects moving in a field of sensor networks. And it has elements of distributed sensing and signal processing, which also involves distribution computation and estimation of states, essentially modeling signal processing and estimation. And we need to go through multiple simulations. Much of it is currently done by students in a tool like MATLAB itself. I'll expect to see if we can put in proper use of Sylab for this. I'll be very happy to encourage that. So the problem is like tracking an object moving through a sensor network like this. So each sensor collects some data about the tracked object and transmits to a base station. And there is an estimation tracking filter in a base station which computes all this. So basically we model the system through its state space approach, look at its motion. And since this is a sensor real network, the sensing, sense data is often corrupted with considerable amount of noise. So we need to filter that out and come up with good estimates. So this is a statistical process, stochastic process wherein we bring in factors about the signal and its noise which is modeled through probability distribution assuming Gaussian distribution there. We have a model of the measurement model of the system which is typically incorporated in the form of matrices like this. And we come up with the prediction of the state of the system through a set of operations in sequence. This is a very simple sequence we use here is the Kalman filter. I'll not go through this in details, any references. But I'll just give you how we get to use this. So typically when we try to do a simulation, a simulation looks like if a object is moving in straight line through the sensor network which is spread out densely, then you'd like to look at how the state, the x position, the y position of it is estimated. How many sensors are involved? What will be the error it is generated in our estimate? And we look at it as it goes through various steps of simulation. So simulations are typically like in straight line motion or a motion which is non-straight line sinusoidal motion like this. Or you go through circular motion then how the sensors are involved and what is the nature of error we get in this. So the objective of this is basically to design efficient codes or algorithms to get good estimates of these positions. Repeated simulations of this are required and such kind of tools are very useful for us. The next one is an example is in autonomous vehicles. We are actually designing an autonomous vehicle for the Indian government. An underwater vehicle which will move on its own steam with its own ability to navigate itself. So basically if a navigation has two parts it has to localize itself that is identify its own position and then decide what to do next. Much of this design activity is focused on creating a good control scheme which will be implemented in the vehicle and then launched. Several simulations of this have to be done beforehand before we really concretize on it. So typically here the purple element shows and vehicle moving through an area and it localizes itself with respect to certain landmarks and this is what we call a localization and mapping algorithm here. One can do it in the same way like we do a Kalman filter but this is an extended Kalman filter where when we use a vehicle model and a landmark model and we use the sensor model with its noise observation captured in it and you recursively apply an algorithm to predict the state of it. Quickly go through this and we can use applications the sensors like variant sensors with a inertial navigation system and a compass and taking its real data of actual vehicle systems and then look at how an object will move. Like this is a typical simulated output of a vehicle trying to move in a circle and it doesn't know much about the environment. So it takes a few landmarks. The green stars are some of the landmarks. At point to point of its movement it picks up some landmarks at a time and re-estimates its position. So the actual trajectory is, it's supposed to move in a circle which is there but the actual trajectory deviates from that a little bit because of the noise it picks up and a few landmarks it uses. So the trying to define good algorithms for that is required. So we try to look at micro level at the structure what happens here and see the effect of various sensors, et cetera. So this is where all these curves and data is generated out of a numerical computation toolbox and the codes for this are written by in these toolboxes itself. I'm glad to hear today that Kalman filter applications are coming up in the next version of SILAP. So we'd like to see how we can use it. I'll end this with a quick little more detail about how we use control systems in a another mechanical application that's a gripper which we are designing using compliant mechanisms and these mechanisms are actuated by micro grippers which are actuated by piezoelectric devices. So this is a type of a diagram of it not exactly in scale. This is an earlier model. Typically a gripper which is moved by piezo actuators, it can open and close. It doesn't need any joints inside it, no motors and all that. So these are modeled using mathematical representations. So we simplify this into a equivalent mass spring damper model and then get its equations and this is what we can actually simulate and create a model like this in a simulink. This is how a student has done it in simulink. The same block scheme can be used very easily in psychos. You can see that. The also used tools like sim mechanics, it would be nice to see how we can see sim mechanics equivalence in Sylab or psychos. And sim mechanics allows us to directly represent the mechanical system modules into in the simulink version of it. So here we look at again performance with various things. Like when there is a position control implemented, when there is a force control implemented, et cetera, these kinds of things with all kind of PID controllers built in them. I'll conclude with this. Brief is computational mathematical tools are very useful for us and there's no doubt about it if these are available in free domain, I'm sure many users will come in to use it. MATLAB is perhaps more prevalently used right now amongst the IIT students. So it is proprietary, so we have dependence on it. We'd like to move out of it soon. Use of open source tools like in Sylab should be able to extend the innovation in education as well as the design processes here. So if you really want to have Sylab being used, we have to actually think of ways of getting industry also being involved in this process. Maybe you need to have a version of it which is like priced and sold and sold as a product, supported as a product also. Like MATLAB, I'm not sure about its economic viability, et cetera, but then this is a reality. I would say that in projects where there's an end use of the research which is going out, they want deliverables in a code which will be run by anybody in the company, not necessarily people who are graduates from these IITs or other institutions.