Thank you very much for Professor Hutchinson. Look bad army because our local I'm competing with the young people for a seat guy. And so by me, thank you very much for inviting me to give an overview because I was asked to give an overview. So what I do is I go to Spain. My talk into three parts, okay, So I will be divided into say, five no more than five minutes about talking about quick thing and then in five minutes about something related to what we're doing. And also another five minutes for the whole robotics. Now, just a quick comment on it, okay? Because I'm here on the other spectrum. I'm here at 40 years. And professor Hutchinson say that while it should be robotic incentive and get rid of the intelligent machine. But honestly, if you get rid of intervention machine, I don't have a lot of role in these things. So personally, I think the Intelligent Robotics is an intelligent machine, but not all interdimensional robotics. So that's what you will see. But today I am very focused on few things that how starting with the digital machine patient, the last student in my class while spree, understand what I'm talking about. Then I'll talk about physics base and controlling, design, manufacturing and modelling for direct fuel based control. So these two together, I spend no more than findings and the rest, I will talk about a segment of my work that related to what I'm good at. My title is now the long haul robotics. And actually it's a new area that we're trying to get into it. Now, usually what I like to give you a perspective of what my research, how I select my research. So I'm here in 85, but I came to this country in 19.5781 year in my undergraduate at MIT. And the rest of my time, lag time is here. Now you see that when I was a graduate student, just like you is actually at my process. And then when I Georgia Tech is information highway, and then you see you see those gray color things to pay mentioned thing, things that happened. Of course not all of them. Black swan and those kinds of events that direct or indirect actually extracting online. But certainly he had some influence on the weight. I select my application. I see. During the 2001 I worked at that time, I worked a lot with Gary yeah. Marine GTI time. So to be around 911 and side at point for processing become actually very critical area that we try remote human. And of course nowadays you can visually, that same thing happened. And also in-between, we are trading between two trait or midnight AT and 10,010. Sorry, not using my mouse. Now, you see the next level. I'm going to talk about how technology moved during that period of time. Now actually before before 80, before 1980. He was shape right there. I was in at night in the 70 I was in college late seventies. And before there was they came to, I tried to build my, my plus audio amplifier using take him to and transistor very quickly, you go too cheap. And I built my first PC using z eight. Now the point I want to say is that you look at these cover like 2010 to 20 years. But then for maybe five to ten years, seed technology should very, very fast in-between USE that a lot of commuting, communication control and all those things happen very quickly. Now you see we have moving from hardware development, the more software and more knowledge. Okay? And now of course you can see now one particular thing and I say new chances as what I'm mostly focus on. You see, you got to do it. I'm sure a lot of you do work on reputation. One of the major shift key on reputation, especially a stroke patient or literal, which is shift from center base. With therapists do home-based care. And they make, they have a very bad big impact the way that we develop robotics for that complication, That's what I'm focused. But today I'm not going to not planning to talk too much about that. I'll come, I can point of your Summit. The other is more energy. Energy is what we talk about the Double Robotics that Irene fund, the initial part. Now I guess the point I'm trying to say, robotics and machine, you'll see that looking at robotics and other intelligent machine, machine. Now, during the earlier period, the one of the most focus you can see most of the measurements sensor or more point form. Point based on measuring force. Actually. This turn and things like that. Now you see it moved the world. I feel like using acoustic. You're seeing machine vision. But earlier work, at least before I came to Georgia Tech, almost all the cameras are actually for human camera and then just trying to use it for machine. And during that time it's actually $50 thousand per system. And its greatest scale and take about 2.5th to just find a point, a dog in the image. These drugs to actually develop the machine vision. And what I show you here is the one that actually very first-generation or industry vision to actually develop at Georgia Tech, myself and Dr. Dickinson in the lady of course, meet 90. This is not you still can buy these cameras is not a point under sees that. There's a turning point. Wise, my racist concern, moving point, try to point it to the field. So in other way, lumped parameter to Burma. Now that's exactly what I'm trying to show you that when we moving from lumped parameter point-based, not, not moving away by actually expand the lumped parameter approach into distributed approach. And from point-based to feel base measurement. And now we are trying to use it more physics beyond what just like indegree, other kinds of physics media, either side, top and motor control again, most of the machine vision and especially in countries like a lot of time that you're doing a neural net, you either spend a lot time to get the data, to train all you have spent a lot of time in computing. So by our control oriented, we have to do both. Have to have quick way of doing optimization and also a quick way of trying to do in real time. Now, so this just go through a quick thing that we did. Nine cities. A second version of my camera that I invented. Actually we use it for, you'll see that we're using with neural net work that actually, that project actually was motivated by a problem in AT&T down, not cross, because they are trying to keep ten different part thereof on paying one place and using human just like you. But of course, they don't do that anymore because if we don't need desktop phone anymore, so but at that time we're using, we're using 1010 camera pantry and one robot to do it very slow work. And so we propose that we were using tracking moving objects. And that was that location actually see in high base youngsters using it exactly the same and that picture actually say but in ID3 cover page or the proceeding, that was the place that we have. Now then the other one is that we're using machine vision to find the orientation. The orientation of a spherical Moodle. There's physical model basically is a modal. They have 11. You can visualize days, more joy and your ball at one magnetic and they were bundle quite outside. You can actually do the your freedom it even more than two and do the spin. So that's how now you see that we using the same camera that will develop the way to draw the green line. The green line actually is a 2D barcode for the camera to find orientation. Then in the information highway, whenever people think about information highway, always think about that. Computer science and WE, but ME actually doing a manufacturing a play an extremely important role in that. And fiber optics use at that time was produced by ten centimeter gracile and drawing the one at any point micro. So you can visualize this layer hack, right? Done, anything stick onto the glass will break. So in order to do a Control K, traditionally they using the draw my new drove master. You can draw that from one centimeter. Glass rod. But when you have wondered, drawn from a big classroom, so we can kill, that wasn't not kilometer of fiber. You wouldn't be able to do the traditional technology because no longer lump parameter can do because it's distributed. So not only we have to understand the temperature profile, but also we have to understand the grass pro rata. Now, we were at that time, but one of the two organizations that actually manage to be able to solve this problem. And of course, we are close to enough crossed out house so we're able to do a lot of this plan. Now, another one is that we were using the camera. We extend the camera to do there. So we'll try writing. So in other words, the right information into depth into this so that you don't have, you don't need this sensor to do that. So we'll try to do the positioning of them. Now, the point I'm trying to say is that all these, all these, all these control using the existing feel like thermophilic fro field and the magnetic field to do the case. So that's why you don't see sensor inside the fiber optic because you just cannot afford to have a sensor. Because once it is 2 thousand degrees C, That is correct. Stick on the glass, it break. Okay. So we have the using the existing field measure at some point. So what we did is that we're using a thermal camera before I put it in grass and look at the temperature. So we're able to do from single, single input into multiple input. In this case, we can control the temperature of the, of the finance. We can also feed rate after draw speed. So maybe that's how I solve the problem at that time. Okay? Now then this partner, that column is that I work pretty much always GTI and also a University of Georgia and also unit or units are portraying a Association. So is about 15 years of work with Gary. Some of the things that two things here. One is the cutting. I ping when I beat Garry, pay the other party, cedar, live chicken, how they actually look, hang lights, you can add a phosphate. These are again, I have a lot of recognition for that whenever I present this, everybody remember the chickens. So Vinny, thinking about media, always remember this day because I learn a lot from this project. You will see a lot of things that happen after this actually is. Again, a lot from here. One is how to deal with complex hunt deer with a natural object. Okay? And also now here, the machine vision, I'm not planning to talk about. The machine vision play an important role in term of inspection, handling that kind of thing, and locating. Looking ahead, don't have time to talk about. The last one actually is a field K. Let me skip some of these. I'm going to talk through some of the machine perception. Now one of the things that this example I want to show is that the importance of physics in, in, in trying to design machine vision. Now, you'll see here is that it's a touch sensor. It basically a sensor. But we using to stereo cameras. Tiny Tuesday your camera with with with transparent not you're interested. You can look, you look at this just recently published. But most of the time when you're doing this work. And I see that the first method is basically triggered a camera and then we followed that without the stoma. Okay? And the second method is that when you try to do the calibration, you actually treat the electoral polymer as part of that lens, right? So they don't care about a physic. This basically bind it using camera calibration that was developed by length, by psi. I began belong in the 18th. Okay? Now what we did is that we add in the physics, are we adding the physics into the model? So in addition to camera model, we do not explicitly model, we explicitly model the refractive index. So let's take this foot method one. You can see now you do it method one, you'll see that it's actually, you are seeing image, not really a true image. Now the second one is that what people do is that they trick the stomas as part of it. But this failure, you see there's two c, k actually two ways. What other people You're, we introduce some physics. We basically do more Cadbury. We using more different shape up a compliant, compliant rubber as part of it. Now I see again, even those, you can actually do the measurement but actually doesn't give it a physics. But what he did is that we, we account for we account for physics right? In the calibration. So we can see that you can actually get exactly the right, the true feeling. Now. So this is what it looked like. Almost pages. But you see that this object is very small. Touch finger King. I can look at the shape and actually compared to existing method is very, very good. Now, let me jump in and do their last pattern. Okay? All right, let me do a quick thing that this project, now basically what we did, this project was actually working with Professor Chan in WE and also with a petrol company. Now the whole idea is you're trying to design a robot then going vertically down about one thousand, one thousand meter and be able to carry some sensor things either timeless, give about them. Now, basically we choose the telescopic type. Not a key idea is that we do not want to have any external sense except though miniature magnetic sensor. And also we measure directly the magnetic field. So we instead have to carry a lot of a large battery. We're using a lot of small miniature magnet. So as, as energy shells. And there are two main part of care. One is a complaint beam and the other one is actually HIPAA. Alright. Okay. So let me let me just the one interrupt you here. Scheduled and we're very tight. You can speak with. I encourage you to meet one another, but also not too late.