Sarah
Welcome to a new episode on the STEMterviews. Today we have with us Dr. Krishna Kidambi from the University of Maryland, where he’s currently working as a postdoctoral research associate in the perception and robotics group. Welcome.

Krishna
Thank you.

Sarah
So just as a beginning, can you maybe just summarize your research project for our audience?

Krishna
So currently, in my postdoc, I’m working with the perception and robotics group. I have more of a theoretical mathematical background, plus controls background. So in our group, I work on specifically controls and estimation so how do we control a robot, or like how do we guide a robot to do a specific task. And also I write mathematical equations that say, okay, this is how the robot should behave. And the perception team says okay, this is how the sensors behave. So I kind of integrate both of them to efficiently navigate a robot indoors or outdoors. It depends on the application.

Sarah
Okay, so what does perception mean for a robot, in robot terms now?

Krishna
Simply, equivalent to our eyes. Perception is how do you see the environment. That’s basically perception or vision. But there are other technicalities, like sensors that we can associate perception with the sense of touch. But I predominantly work on vision.

Sarah
Okay, so basically, you teach a robot how to see?

Krishna
Yes. Not teach per se, but how do we infer what it is seeing. And how do we effectively…it’s basically ones and zeros or some kind of random information. How do we tell the robot, “Hey, there is a cup here or there is a human being here or there is a door here?” So that kind of information. We have to generalise it. We cannot just say okay, take this information, find out, navigate. It will not do anything; robots are not that intelligent. So we have to tell them, “Hey, do this, do this” based on the information I get from vision.

Sarah
So how do you do it? How do you tell the robot that this is a cup, and this is a tree?

Krishna
There are two classic problems here. It’s very hard to estimate depth. That’s a classic vision problem. So what is depth in the sense, like, how far is the object away from? So we need to…we have two eyes, so we can easily identify the object is one feet away, or I have to get up and move to pick up the object, or I have to avoid this. But, classical problems are such a way that you need two cameras to identify. And the problem becomes whether you’re indoor or outdoor. So if you’re indoors, then there are other sensors that I can use, and I can say, “Hey, robot, you’re here, go to a point B”, which is this distance away. So I know, at each and every point of time where the robot is with respect to the real world.

Sarah
Because you have other sensors within the room or wherever.

Krishna
Exactly. But there are others. If you come outside, there are GPS and other kinds of information. But sometimes you lose GPS information based on because of cloudy cloudiness, or you might not have GPS satellites, where you are on top of you. So in that sense, a robot doesn’t. If you try to close your eyes, and then try to walk in a 3D environment, you would not go, like, you don’t know where you’re going. So that’s useless. I can walk, but what am I doing? Where am I going? There is no inference from that. So that’s probably localization. So those are two classical problems of vision.

Sarah
Okay, so why do we need a robot that knows that there’s a cup outside? Outside, not just inside? Are we building robots that are supposed to work outside?

Krishna
Yes, like that. That’s another application, right? Like a lawnmower example. You don’t have to have a person mowing the lawn. But at the same time, if there is somebody and if the lawnmower encounters a person, then it has to stop, right? It should know how far is this person.

Sarah
Please don’t kill this dog. This is a dog by the way.

Krishna
No worries. The robot is never going to come outside again if it hits a human being. So I think like that, how far is the human so it has to analyze it, and then like, “Hey, I am here. The human is 5 feet away from me. So slow down, and stop if I’m like, closer to him.” So that kind of information is needed for the robot. Like, do anything. I gave you one example. There are multiple examples.

Sarah
But isn’t it like, don’t we have self driving cars that are kind of doing this already or learning this behaviour already? Or am I wrong?

Krishna
Yes, there are. You’re perfectly true. But they have multiple information, like, multiple sensors. There might be sensors in the front, they might be, again, as I said, they might have LIDARs. They are like, they scan the 3D space and then they give you whatever the distance to objects. So again, one example I can give you is the Tesla car because it has so many sensors, and it has so much of computation in the car. So then we can do. But if let’s say, I cannot put the LIDAR on a small lawnmower. The lawnmower is, like, $200 and the LIDARs might be, like, $2,000. So this is a feeling you have to take into account.

Sarah
Okay, so this kind of explains or already answers the question, why is the project important for our audience. So how, what’s the application other than a lawnmower or a car? Where else can we use this project in our daily lives?

Krishna
Okay, I’ll just deviate. I come inside now. Let’s say you have a robot that’s helping an old age home…old people in their houses. So how does it pick and place the dishes in the dishwasher? That’s an example. It has to move in such a way I don’t hit the person. And I don’t break the object. If I’m picking, let’s say this cup, I’m picking the cup. I know where the cup is. I know what is in it. And how much of it. I can do all these calculations in my head within a fraction of a second. If I try to pick the object, if the object is glass, if the object is plastic, so there are two different kinds of forces involuntarily applied like my hands apply that we don’t see. But we have to tell the robot, “Hey this is a cup. And this is a glass cup. So you have to pick it up”. If I crush, like, if I put more force on my cup, it breaks eventually, which is not good. So you need to identify all these things. That’s indoors again.

Sarah
That’s indoors. Okay. So how does it work? Are you yourself applying for grants? How do you convince grant givers or non scientists of your research? Why do you think it’s worth wondering?

Krishna
I think it would help. There are, as I again said, there are multiple applications. And then we tell, “Hey, this is what has been done, and how can we improve this? And this is what we are going to do.” So that’s how like, what has been. That’s how grant proposals are funded, right? What is new or innovative about this? Even though it might not be feasible, but the idea is innovative. So like, “Hey, can we do this? We don’t have money, give us money, and then we’ll try to solve this problem.”

Sarah
Yeah. So where are the limitations then right now? What kind of grant are you writing for at the moment, or was the last grant you wrote?

Krishna
I think, just after my PhD, I wrote a grant. And then it got funded. But I had to wait for some time. It’s like, but then I was okay, I cannot wait. So I then picked up this position. So then I wrote…we’re writing a grant currently. So that says how does safety come into picture.

Sarah
Okay, that’s interesting and super important. Tell me more about this.

Krishna
So, right now we are working on how do we make robots safer. Like, how do we make machine learning safer in very simple applications, not like complex applications. Let’s say you’re navigating or let’s say you have some kind of sensor, and it’s giving you noisy data. How do we use machine learning to identify that noise and eliminate it before we give into…before we take some next steps? So those kinds of…I’m writing that grant, like, how does it prove it safe, and at the same time, we use good signals for control.

Sarah
Okay, so what’s the issue with noise and safety? I don’t get that link.

Krishna
Let’s say if your system is noisy. Or okay, I can give you a simple example. Like, I have my glasses on. So if I try to remove my glasses and then see some objects, I might not be able to see them clearly. It will probably be hard. Then how efficient my brain is? I am still doing the mundane tasks ineffectively. So then if I try to correct my vision using glasses, then I see the object clearly. And then my control system is saying, “Okay, pick up the bug. It’s 1 feet away from you.” So then I have good information from sensors. So that’s how sensor noise is like very…let’s say, even in small vehicles, let’s say, the lawnmower I gave you example, if I have so much noisy data, I would not be able to perform properly the mowing of lawn.

Sarah
Okay, so with noise, you don’t actually mean sounds, you mean like the amount of data that you have that don’t give you specific results, right?

Krishna
Exactly. I might have data, but that’s useless to me. So, how do I get the useful data from useless or like, some kind of information, extract information from the noisy signal. Noise in the sense disturbance, I would say. Like, you’re hearing music and stuff. When you have a loud noise and stuff, or there might be disturbance, there’s a persistence. So that’s basically noise.

Sarah
Okay. Just for our audience to know what scientific speaking noise can be. And it’s not all about sound.

Krishna
It’s not all about sound.

Sarah
Okay. So one of the questions from our audience actually was, how do you see the future of robotics? What do you think a robot could help you in, I don’t know, 20-30 years time?

Krishna
I am actually too young to predict, but I can give you my inputs to see what.

Sarah
What about the lawnmower?

Krishna
I think there are actually those kinds of lawnmowers inside the house, right? Like, we have the Roomba. We’re already there. Have you heard of the Roomba? It’s like a vacuum cleaner inside the house that goes. When there is an obstacle, it tries to turn around. It’s very simple.

Sarah
Yeah, I think most people know, yes.

Krishna
But the direction we go in maybe…how do we again, as I said, if the robot is interacting with a human, how do we account for safety? And also…I had something in my hand, suddenly it slipped. Just give me a second.

Sarah
No worries

Krishna
Yeah, I think human-machine interaction plays a bigger role as we go along.

Sarah
Okay, that sounds really interesting. So how civilized robots are going to be to a human in the end. Are we going to have a proper conversation with robots?

Krishna
You still…you have that already. iPhones…

Sarah
Okay, we can call Siri. Yeah. I know.

Krishna
But again, if we have to go into let’s say, each house or we have cell phones in each household, it might gonna take a lot of time. We are nowhere close yet.

Sarah
Okay, that’s quite relieving for me, to be honest

Krishna
We have some kind of robot manipulators. They’re called…your hand is a manipulator. So it moves freely, right? And your vision is saying, “Okay, there is a wood, don’t go in that direction because you’ll hit.” You have that prebuilt information. Can we teach all the 20 years or like 25 years, what we have learned to a robot in, like, next two, three years or next five years? Probably not, yet. We might, we are working on that.

Sarah
Okay. So what are your feelings about the ethical issues that we have with robotics and artificial intelligence? For myself, I, for example, don’t like talking to Siri or any of it. So that’s why it doesn’t even occur to me I’m not having a conversation with a robot.

Krishna
That’s a very good question because I had this discussion with my wife. Let’s say, if somebody asked you to build a robot that can be deployed in a war or something. And I was, no, I’m not going to do an algorithm that does it. So I think it’s individuality becomes a bigger issue there. Somebody might be like, “Hey, this is my technology. I want to develop it, it doesn’t matter where I use it for.” But again, then regulation should come into picture, okay, you can use under these circumstances. There should be extensive regulation saying that it cannot be used, I think. I don’t know if they have that, but some countries have used. They have a treaty saying that you cannot use AI in warfare, or I’m not sure, or maybe I’ve seen in a series or something, I don’t know. So it’s going to be a big, big thing like as we…because we cannot teach consciousness to a robot yet. I don’t know if it’s possible. I think Einstein, no, Stephen Hawking said it’s gonna be ‘really cataclysm-istic’, or it’s gonna be extinction level event if we teach consciousness or intelligence to a robot.

Sarah
Because then we’re completely substitutable.

Krishna
Yeah, that’s what it is. We are on the top of the human chain, like, the top of the food chain. So if we teach a robot intelligence, and then they are on top of the food chain, so you’re not the alpha in the food chain.

Sarah
Yeah. Okay. Let’s not do that then, I guess.

Krishna
Hopefully, but you never know scientists sometimes are mad intelligent when it comes to real world scenarios.

Sarah
Hardwired.

Krishna
I guess it’s, they have written on. I think predominantly, that experience comes from, for me at least, from movies, where I found, “Hey, that is really stupid. Why did you even do it?”

Sarah
Then, I mean, who wrote those movies? Not scientists probably.

Krishna
I agree. That’s a good point.

Sarah
Okay, so you said you made quite a break from…you shifted your interest after your PhD. So what was your PhD project about then?

Krishna
Okay, simply put, it was about how do we regulate the flow around an aircraft. Flow is like air. Flow is some kind of fluid moving in through a medium. So my problem was…you have flown, right?

Sarah
Yeah.

Krishna
So there is…an aircraft has certain control surfaces they’re called, like, how does its lift happen? How does it rotate? Like, let’s call, roll and, like, how does it yaw, stuff like that. Traditionally, control algorithms, like, aircraft control is done using some mechanical surfaces. It’s a metal that goes up and down so that you get lift. The next technology is like, they’re called synthetic jets. They are placed on a wing. And then they release a puff of air. So it modifies the airflow. And then you can orient, you can make the aircraft go crazy manoeuvres at high speeds like supersonic. Not enough to a passenger flight, but like an unmanned flight. How do we go supersonic? How do we go hypersonic?

Sarah
So how does it work? Tell me. How does a puff of air make a jet go crazy, as you said.

Krishna
Okay. Simply put, I want to rotate the aircraft or I want to change its direction. At high speeds, if you try to use these control surfaces, as I said, metallic surfaces or metals or some kind of control surfaces, they will not be able to deliver your design performance because they’ll melt or they deform at those high speeds. So instead of these traditional surfaces, we are using synthetic jets, they are like some kind of membrane. Like, if you give a voltage, it releases a puff of air.

Sarah
What sorry?

Krishna
A puff of air. So if you put them along the wing boundary, and then if you activate them, and then you can bend the aircraft faster, or you can rotate, I should not say bend, but rotate or do whatever design **inaudible** in an efficient way, because there is no mechanical surface that’s being actuated. It’s easier, and it’s more efficient. And I provided theoretical guarantees. I would say, the control engineers are more interested towards it. I mean to say okay, this kind of work can be done. I gave mathematical proof saying that this is stable.

Sarah
So you basically just took the theories and the physics behind it and said this is what the model says. Yes, you can build this aircraft?

Krishna
Yes.

Sarah
Okay nice. And is it gonna be built and when? Before or after the robots?

Krishna
That I don’t know yet. But NASA is working on a…it’s called X aircraft, some kind of configuration. All NASA aircrafts are labelled X if they are demonstrative aircrafts. So NASA is working on a project, but I don’t know yet what is the name of the project. I’ve seen some images, I think, I don’t know if I can. I’ve seen some images where it looks like a, how do I put it, it looks like a horizontal. No, that’s not a good example. It’s not a traditional aircraft to be put simply. It’s like a hypersonic or supersonic aircraft.

Sarah
Oh my god, I can’t wait to see a thing.

Krishna
We’re far away from that as well.

Sarah
Is it going to happen before or after the robots? What do you think? What’s your take on that?

Krishna
Robots are already here.

Sarah
We also already have aeroplanes. We’re fine tuning now?

Krishna
I agree. Yeah. Maybe, before we expect robots in our home.

Sarah
Okay. Awesome. So I was also told that you’re interested in science communication. More or less?

Krishna
Yeah, because I have my PhD in sciences, or my degree comes from the College of Arts and Science. So I am like, we always debate I’m an engineer or I’m a scientist. So, yeah it’s, you can say….

Sarah
…something in between. Okay. So what is, why are you so passionate about science communication?

Krishna
I think this is the big thing that I realised after, I think, my bachelor’s because I didn’t pay attention to the basic sciences like, classical mechanics or physics. Early, early stages of the base, I should say, was not set properly for me. So I think I really…

Sarah
So you’re doing all this physical modelling without knowing the basics of physics on fly in this aeroplane.

Krishna
Oh no. After my bachelor’s, I realised, okay, the ground should be solid. So that’s where I entered into the basics of sciences, physics and stuff. So I somehow I have a connection. I teach my kids like, “Hey” because I taught physics, like, entry level physics to engineers. So I told them, “Hey, I didn’t do it so this will haunt you to the next 5 months, forget about 5, the whole rest of your life if you want to become an engineer.” That’s how I became passionate about basic sciences. And I like to teach them from the ground up. If your basics are strong, you can do whatever you want to go up to. And one of my students came to me and said, “Hey, Krishna, you told me in my first year that this class is important. And I am still using these equations in my third year or fourth year of my bachelor’s degree.”

Sarah
That’s important. Awesome. So what are your favourite science communication projects that you’re involved in, other than teaching your kids about the basic science, which obviously, is the most amazing one, most important?

Krishna
Currently not. But I might ask you, I have to ask you a question, what are science communication projects? Is there a specific term for it? Or I don’t know.

Sarah
Like anything outreach or public engagement. Anything like that.

Krishna
Oh yeah, I have, we have done something like…there were high school kids here around this campus. And I’m associated with the Midland Robotics Center. So they came to us and we showed them, “Hey, these are the labs and stuff.” And they are, actually frank, way ahead of how when I was at 16 or 17. They are building robots. They’re writing code and like, they’re doing stuff, which I did at 23 or 24. I still. So I think the kids here are really advanced, or they know more information maybe because of the internet.

Sarah
Wow. Amazing.

Krishna
And also, I volunteered at some high schools for I think when I was doing my PhD. I was a substitute teacher. I just go with not the high school kids, but like the primary kids, see how the system is like, help them out. So I’ve done that work as well.

Sarah
Okay. That’s good

Krishna
A few things did not materialize because of COVID. There was a science STEM competition around the area that I live in. And I volunteered for it, “Hey, I’ll be a judge.” But because of COVID I think they cancelled that event. May 19 or early, early 2020. Sorry, I lost track of 2020 because of COVID.

Sarah
Don’t be. Oh, my god. Okay. And one really important question from our audience. Because you work in such a probably men. No, okay, let me rephrase that. Because you’re working in a field, which is probably full of men, are there actually lots of women in the robotics group or in the control engineering field that you’re in?

Krishna Yes and no.

Sarah
Yes and no?

Krishna
Okay, the reason no being like I’m very…it’s very sparse. It’s very limited as well. And I said yes because my wife is a controls engineer. And I have two of my PIs who are women. And also, when I joined postdoc here, out of the four, two of them were women. So it’s growing at the same time, but it’s not as rapid as it should be. But there are…now there is a society for women in controls. IEEE has a branch, Women in Control. So I think it’s…

Sarah
I like that name, the Women in Control.

Krishna
Yeah. But there are exceptional, really amazing women who do exceptional work in controls.

Sarah
Nice. Awesome. Now towards the end, we always have a couple of personal, more random questions. Are you ready for these?

Krishna
Oh, yeah, sure.

Sarah
Okay, what was your favourite subject at school?

Krishna
At school, I would say math. During my PhD, I would say science or physics.

Sarah
Okay, that’s sticking with one side. One sentence: What are you truly passionate about?

Krishna
I’ll say teaching the younger kids.

Sarah
Okay. Teaching so not just your own kids, but students.

Krishna
Yes, like the high school students or like early, even below 10 or…

Sarah
Oh wow.

Krishna
…10 to 15 or 10 to 18.

Sarah
They’re probably super keen to hear about robots and high speed airplanes, right?

Krishna
Not technically, per se, but I love teaching math. And numbers are like, they should like, I didn’t have this. My dad was a teacher, but he was exceptional. I didn’t realise it at that time because he was my teacher. So I used to be scared of it. But he was later, I realised, he’s exceptional teaching math. So that’s why I think I have that interest of teaching from both of my parents.

Sarah
That’s good. Okay, what are you doing in your free time when you don’t teach your kids science and you’re not in the lab yourself? What do you actually do?

Krishna
I hang out with my friends a lot. My wife complains that I have too many friends.

Sarah
Okay, that’s unusual to hear.

Krishna
I like cooking. I explore if I’m in a town. The first thing I do extensive research on what are the restaurants in here? What are the foods I have to try? So yeah, I like eating and cooking.

Sarah
Nice. What is your favourite movie including robots?

Krishna
Oh, wow. Including you want the movies with robots?

Sarah
Yeah, your favourite movie that has robots in it?

Krishna
I cannot pick one. But I can give you a couple of names or two or three. Maybe. Okay. What is it called? I like Real Steel. It’s a movie about robot boxing. It’s really fascinating. And I think there is another movie from the UK. It’s called Ex Machina.

Sarah
Ex Machina.

Krishna
And maybe the I, Robot from Will Smith.

Sarah
Yeah, I like that one.

Krishna
Yeah, they’re like, you can see how a robot takes over. That was what, 20-15 years ago, I think.

Sarah
Oh god. I feel so old now.

Krishna
Yeah, I, robot is one.

Sarah
Okay, and that might be really interesting. But what would you do if you were donated $10 million to your project?

Krishna
Oh wow. That’s a tough question for a project. Okay. If they are giving me $10 million, it will be at least for 5-6 years or like, next 10 years.

Sarah
Yeah, but which problems would you take over? Which challenges would you want to try to tackle?

Krishna
Specially building a robot is difficult. That’s what I’ve learned here. Because like doing mathematics and like theory, research, it’s easy. It’s just have to say, you just have to say, read some papers, work it out. But building a robot…the robot you build today will work today. Sometimes it might not work tomorrow. There might be n-number of problems like, this wire has gone out or this motor has gone out or this sensor is faulty. So I don’t know, maybe efficient building I would focus on like, how do we build robots that are efficient.

Sarah
And sustainable as well.

Krishna
Yes, exactly. And safety again. As a controls engineer, I worry about safety. I would put safety predominantly into the project.

Sarah
Yeah, that’s important to tackle. Okay. Thank you so much for this nice conversation, for teaching me about robots and all these interesting things.

Thank you for watching with STEMcognito. Find more videos using the search box or the drop down menus above. If you think there’s something wrong with this video, please use the report button to inform the STEMcognito team. Questions about the video content should be directed to the researcher. You can find their details below. Go to our submission pages to find out how to submit your own video and don’t forget to follow us on social media.