Oral-History:Robin Murphy
About Robin Murphy
Born in Mobile, Alabama in the middle of the space race, Robin Murphy and her parents would move to Douglas, Georgia where she would be raised and attend school. Murphy was greatly influenced by her father and although many people kept asking her about the “new stuff called ‘computers’”, she decided to follow in her father’s footsteps and would attempt to earn a degree in mechanical engineering. She attended Georgia Tech and received her undergraduate degree in mechanical engineering. Robin Murphy ended up working for many different companies like DOW Chemical and Turbitrol (Now part of Johnson Controls) until she realized that she did not enjoy what she was doing and decided to go back to school. Because of the work she was doing with computers for process control, she decided to go back to Georgia Tech and get a master’s degree in computer science and immediately fell in love with it, graduating with a masters in 1988 and a Ph.D. in 1992. Murphy was the first person to ever graduate from the Georgia Tech College of Computing with a Ph.D. in robotics.
Upon graduating, Murphy began her teaching career starting as an associate professor at the Colorado School of Mines from 1992-1998, moving onto the University of South Florida from 1998-2003. She was promoted to full professor in the year 2003. From 2003 to 2008 she would remain working at the University of South Florida until she started working at Texas A&M University, where she remains today. Robin Murphy is well known for her work on disaster robots that date back to 1995, which were motivated by the Oklahoma City Bombing. Her research is looking into how robots can help us through times of disaster. Her robots were the first to be used for the emergency response phase of a disaster when the 9/11 World Trade Disaster occurred in 2001. Since then, Murphy has helped insert unmanned ground, aerial, and marine systems into 27 different disasters including Hurricane Katrina in 2005, the Fukishima Nuclear Accident of 2011, the Tohoku tsunami in 2011, the Syrian Boat Refugee Crisis in 2016, and Hurricane Harvey in 2017. Murphy was also a member of the Defense Science Study Group from 1997-1998 which would help lead to her involvement on a multitude of science boards including the Defense Science Board and the U.S Air Force Scientific Advisory Board.
Robin Murphy has received many awards and honors through the years including; the U.S Air Force Exemplary Civilian Service Award in 2005 for her work on government boards, she was featured in the documentary Living With Robots that was shown at the 2010 Sundance Film Festival, she was awarded the Association for Computing Machinery Eugene L. Lawler Award for Humanitarian Contributions within Computer Science and Informatics in 2014, an honorable mention in the 2015 American Publishers Awards for her book Disaster Robotics, and finally, Murphy is frequently cited in WIRED Magazine “AlphaGeek” as one of the 15 innovators reshaping Texas and as one of the most influential women in technology.
About the Interview
ROBIN MURPHY: An Interview Conducted by Selma Šabanović, IEEE History Center, 22 January 2015
Interview #812 for Indiana University and the IEEE History Center, The Institute of Electrical and Electronics Engineers, Inc.
Copyright Statement
This manuscript is being made available for research purposes only. All literary rights in the manuscript, including the right to publish, are reserved to Indiana University and to the IEEE History Center. No part of the manuscript may be quoted for publication without the written permission of the Director of IEEE History Center.
Request for permission to quote for publication should be addressed to the IEEE History Center Oral History Program, IEEE History Center, 445 Hoes Lane, Piscataway, NJ 08854 USA or ieee-history@ieee.org. It should include identification of the specific passages to be quoted, anticipated use of the passages, and identification of the user. Inquiries concerning the original video recording should be sent to Professor Selma Šabanović, selmas@indiana.edu.
It is recommended that this oral history be cited as follows:
Robin Murphy, an oral history conducted in 2015 by Selma Šabanović, Indiana University, Bloomington Indiana, for Indiana University and the IEEE.
Interview
Interviewee: Robin Murphy
Interviewer: Selma Šabanović
Date: January 22, 2015
Location: College Station, TX
Background and Education
Šabanović:
So could you tell us where you were born?
Murphy:
I was born in Mobile, Alabama, during the space race. I was born just a couple of days after the Sputnik satellite was launched, and then everything in education changed to make kids into engineers, to get them into space, to, you know, deal with that. And so it was a really exciting time. I grew up always expecting that I was gonna be working on a space station, one of those big wagon wheeled ones like you see in “2001.”
Early Work life
Šabanović:
And so did you kind of do things related to that when you were younger, or how did that get you into robotics, or into university, and in particular grader truck?
Murphy:
As a kid I always knew I was going to be a mechanical engineer. Later on in life, people kept saying, “What about this new stuff called ‘computers?’ Don’t you want to be an electrical engineer? Don’t you want to be, you know, a computer engineer?” I said, “No. I’m gonna be a mechanical engineer. My dad was a mechanical engineer. That’s a big thing for the space program. It’s a great degree,” this and this.
So, of course, many years later I wound up with a PhD in computer science, but there you have it.
Returning to Georgia Tech
Šabanović:
So how did you end up in computer science?
Murphy:
I fulfilled my goals in becoming a mechanical engineer. I went to Georgia Tech, fantastic education, and I never did anything with robotics. I had read of them in science fiction and accepted that, but mechanical engineering robots are pretty stupid, you know, their automation. They were pick and place robots, not very exciting. I worked in industry for many years and then decided to come back, because so much of what I was doing was in computer science, using computers for process control, that we, living in Atlanta, I thought I’d go back and get a master’s in computer science, and then I’d be a consultant. I fell in love with it. I realized I loved computer science, probably should have listened to all my friends who said, “You’re actually good at this,” and done that and just never looked back.
Šabanović:
Who did you work for before you went into your PhD?
Murphy:
I worked for several companies, Dow Chemical, as a mechanical engineer doing process control, and then a company in Atlanta called “Turbitrol,” that’s now part of Johnson Controls.
Šabanović:
What did you end up loving about computer science?
Murphy:
I just loved everything about computer science, but what really made the huge difference for me was the idea of artificial intelligence, which up unto that point I had thought was just kind of, yeah, sure, you know, and the same thing with robots. You know, from mechanical engineering they’re stupid. Why would anybody care? There just wasn’t much you could do with that. Well, artificial intelligence is just fascinating. You’re bringing in insights from biology, and cognitive science. How are you putting that together into representations that a silicon type of intelligence can handle and work with, really forces you to think about things in a very broad perspective, and then apply them. So this is just very cool. And then with robots you could make things that actually moved, actually did something that effected change in the world. What’s there not to like?
Šabanović:
So what were you working on during your PhD?
Murphy:
So during my PhD, my topic was on sensor fusion. The time, this was during the late ‘80s, early ‘90s. We’d had the revolution of reactive robots. We were getting robots of different sizes to move using animal like behaviors. So what was the next thing, you know. We have the motor control down. The next challenge was perception, and so that’s what typically has been my career has been, okay, we’re kind of getting it together in this area. What’s the next big thing? So this was the idea of how can you sense better? How can you take lots of okay, noisy sensors and how do we put them together? So going in the classic AI tradition, I went back and looked at the cognitive and behavioral foundations. We know about the nerve physiology of how our senses come together in our brain about here, how it’s shared among different parts of the brain, how different senses, how we combine them together, how sometimes you are looking at something but it’s really the sound that triggers stuff. All of these different things got through that and began to pull that together. And so I still work in sensing as a key to autonomy. It’s hard to be smart if you can’t really understand or comprehend, or react to the world around you, but my work has now shifted more and more to that interaction between the human and the robots, how humans use them, how each complements the other, particularly in a lot of cases, the human is supplying a lot of the more sophisticated perception, because we’re still very good at it, which just kind of makes sense, because so much of this part of the brain is actually associated with perception, and we use the neural circuitry for perception for just thinking about things, for visualizing problems, so it’s a very fundamental to us. The rule of thumb is if it’s easy for a human, it’s hard for a computer. If it’s hard for a computer, it’s, you know, easy for a human. So that things are easy for a computers, like the optimal path pointing that, you know, you wouldn’t even think that way. But perception’s something like, “Oh, look, that’s a cup. That’s my cup,” really hard, and so some of the things we work with.
Ph.D. Thesis Project
Šabanović:
And so were you working with robots while you were doing your PhD, or did you not start that at that point?
Murphy:
So when I started my PhD, I started as a master’s, and then began really just to love the whole idea of computer science more and more, applied for a PhD program, and I was put in something called “Computer Integrative Manufacturing Systems Fellowship.” And you had to do something. You had to work for a professor in a department there. We had one that specialized in pattern recognition, a lot of math.
He was a terrific professor, but just not me. We got this new guy in, Ron Arkin, one of the leaders in the field, and he’s talking about robots. And I’m thinking, “This is just total dreck. I mean, you know, robots are stupid. Robots are that. But, you know, it can’t be worse than this other professor than having to be in this torture of matrix multiplication stuff that I’d never get my way out of. And so I asked if I could work for him, and I became his first student, his first grad student, his first PhD to graduate, and just fell in love with it. I mean, after a month, he showed me what he was learning, what he had learned for his thesis, all the things he was working on, and it was fascinating. It was possible. It wasn’t the mechanical engineering, uh, uh, uh, uh, uh, uh. It was all about real intelligence, and I was totally wrong. I was totally wrong about robots. It was great. It was wonderful, and that’s where I’ve been ever since.
Šabanović:
Can you tell us a little more about the specifics of what he showed you that kind of convinced you basically?
Murphy:
Well, going back, his work, as was most of Rod Brooks, David Peyton, Ron Arkin, really the leaders in this reactive movement, where you’re going back to the lower levels of intelligence and building up layers or competences, these ideas of behaviors, and very few lines of code could get tremendous results, like moving toward things, moving away, avoiding all the analogies. We come out of the tradition of Michael Arbib, of understanding neurophysiology. Seeing how that mirrored what we know about the way the brain’s organized, brains in lower animals, as well, just totally fascinating.
Šabanović:
So did you work with robots with Ron?
Murphy:
That was when I started working with the robots. He got his first Denny DRV. We were there. We started off just moving, how would it move intelligently to dock with something. So you think of a robot; okay, I need to get right there. Well, people kind of do that kind of curve, you know, kind of move in there. It turns out animals, us, we do things. We do usually in two motions, one is this very ballistic, I get close enough, and then when I start getting there, I actually change the perception rate. I change the frequency. I change what I’m looking for and I start paying attention. So if I’m reaching for a cup, kind of like, now I start turning right about the time, and then when I touch it, I actually swap. We swap from vision to being tactically controlled on how we’re doing that. So those types of things were the things I was working on. How do you move into something that’s smooth and graceful and is also energy efficient? And to be able to do it in less than a hundred lines of code, how amazing is that? You know, it was just really, truly a paradigm shift from what I had been taught as a mechanical engineer, where you had to map out the entire world. You had to do that. It was like oh, my God. This really appealed to me. And the two worlds, there are times when you want to use the mechanical engineering. There’s time when you want to use the artificial intelligence methods, and so knowing when to use those, but it was nice having both.
Šabanović:
Did you, during that time, have any either collaborations or contact with people who were doing more of the cognitive research, as well, since you were so inspired by it?
Murphy:
On my PhD thesis, I was really fortunate. I was a Rockwell International Fellow, so they gave me a large stipend in travel money, and I was able to bring in Herb Pick, from University of Minnesota, a cognitive psychologist who specializes in sensor fusion. I mean, it sounds kind of funny, but he wrote one of the fundamental papers on the fact that you know how when you’re in a hurry, you can’t find your keys, and then when you calm down a minute later, you kind of realize you’ve walked past ‘em, like, three times, and this idea that your expectations, we see a huge amount of the world, and we can’t process it all, you know. We think we do. No. That’s not true. What we’re doing is that we’re using expectations. Okay, the key should be where I left ‘em, which was, of course, over here, so even though it’s the light, the physics that’s hitting our retina that it’s there, but the brain’s going, “I can’t see anything. I can only see if you’re over there,” and then how that all works. So he was one of my outside committee members, so that was really exciting. And also to be able to go back through the recent exciting work coming out of the neurophysiology groups, and how cats did sensor fusion, including how they cue up. If they hear something, they start looking, and they’ll wait a certain time until they can catch the motion. Then after a while, they’ll drift off, okay, and go back to what they’re doing. And so that temporal interaction is also fascinating. So, you know, at that point, up to that point sensor fusion had been looked at as, well, you’ve got to build the sensor, and they’ve got to overlap, and they’ve got to have the same signals, and you’ve got to kind of sum the signals together, versus this idea that well, if we hear something with this, then we should be able to see something with this one. But it turns out it’s a false positive. Over time I’m gonna lose interest and go back to what I was doing. So those are the things that I got to work on, and still do to a certain degree.
Post Ph.D.
Šabanović:
And so what happened after your PhD?
Murphy:
After my PhD I became a professor. I’ve been a professor ever since, and the big shift in my career came in 1995, so in the late ‘80s, early ‘90s, everybody was looking at small robots. NASA, in particular, was looking at what later became Sojourner, true, where it’s actually looking at three, small, cheap robots. You know, the idea if you’re gonna send a small, cheap robot, maybe send multiple in case one of ‘em breaks. It came up to be one medium size robot, instead of three small ones, but everybody was looking at how small can we go, how intelligent can we go. And, of course, with the communication links, got to be onboard intelligence. So what can you do with that? So it was very exciting. Well, 1995 was a banner year for disasters. We had the Oklahoma City bombing here in the United States. We had the tremendously devastating Kobe earthquake in Japan. And so Satoshi Tadokoro and myself, who had both been working in this area in small robots, and things like that, just said, “Oh, man. This is where the small robots could also be used to go into rubble. At the World, not the World Trade Center, at the Oklahoma City bombing, they had had robots, but they were the big, mechanical, you know, Sumo, bomb squad robots that were all built with huge shielding. You couldn’t even get them to the site, and if they had driven over they would have run the risk. They were so heavy that they might have collapsed and caused somebody underneath in the rubble to be crushed. Small is good. Small is that, and so that became a turning point, and I’ve stayed in rescue robotics ever since.
First Projects and Funding
Šabanović:
And so what were some of the first projects that you did and, you know, where did you get funding for them? How did you get started on them?
Murphy:
We’ve been very fortunate in our funding from the National Science Foundation. It’s been a true process of discovery. This is one of ‘em. The first project we did on rescue robots was probably the example of why I do field robotics now. So being a good scientist, you hypothesize. You try to think of what is a different way of doing it, and we said, “Oh, of course, the way you’d want to do with this little robots, you know, that you would want to put into the rubble at some place in a building in Kobe, the Murrah Building, how are you going to get it there? So we’re gonna build a big robot that’s going to drive it up, and then we’re gonna drop off a daughter robot, a marsupial robot. Then that will be the power supply to the small one will be that and to the other, and you’ll have better communication. They can throw below will mass, all that, and that’ll be good, and drive in there, and stuff. And so we worked on that concept for a while, and we made one, really, really dumb mistake. And this is one thing that professors and researchers, we tend to do. We forget to ask a practitioner. So this was without actually asking the fire rescue teams what they do. So when I moved to University of South Florida, we were very fortunate at that point, wound up meeting the chief for Florida Task Force 3, which is the Tampa-Hillsborough County, and began, right, and they were kind of looking at us like we’re a little crazy. I think this is one of the cases where I’ve always had a lot of women in my group, and I think they looked at us more as the really bright niece who’s a little clueless, and that first, I think if we’d been guys they would have just said, “No, go away. Oh, you know, I’m gonna be out of town for the rest of my life.” But instead, they kept starting to do stuff with us.
And at the same time one of my former graduate students, John Blitch, Colonel Blitch, was at DARPA, and he was doing small robots for the military operations in urban terrain, which was called “The DARPA Tactical Mobile Robots Program,” building small ones. And he was also, he had been part of the teams at the Oklahoma City bombing, the use of small robots for disasters. So that was a big push. But we had been thinking this whole marsupial thing. World Trade Center comes around, the first use of ground robots. We have our small robots. The big robot to carry the small robots were not field hard. You could never use ‘em in a real disaster, so we took up the small ones just because, and the small ones were the really valuable ones. You would never use a marsupial concept for that. It’s just easier to backpack ‘em in there straight. And after that we have said, “Okay. We’re gonna start just with your exercises, and keep our mouths shut, and you tell us what you do, and what you and the dogs can’t do, and we’re gonna work back from there. So that’s been a big thing. So that’s why I do field robotics. You know, in physics scientific reductionism, where you think of the idea and then you break it down into little, itty bitty pieces, and you do each one incrementally, that works great for that, because we know a lot about physics. I mean, we’ve been around gravity quite a while, you know. We’ve had Newton, and Galileo. Here robots are new, and unmanned systems are new, so we don’t know how to apply them yet, so we need to figure out what the domain is first, and let that cause us to have the fundamental research questions and go from there, and the silliest things you can get out of field work.
We were doing something with Virginia Two, was doing a training exercise with 'em, showing how to use these small-- Virginia Task Force 2, a FEMA team. Here’s how you use the small robots, and this, and this. And I turned my back on that. They were doing a very confined space reaching where you have to crawl in, and breach a hole in this concrete, and the guy’s all in there, and he’s got acetylene torches in there. And they wanted to put the robot in just to take advantage of, you know, before that, because there was no one. Well, sure as heck, the guy driving it had managed to not only get it over this hump, but he managed to get right next to the guy who’s working. Great. Now I’ve got this-- they’re gonna get mad. And instead they’re talking to each other, and the guy who’s working said, “Oh, hey, great. Hey. Can you tell-- can he look at this? Get him in. Can he tell me what’s going on here,” and this. So they start talking to the robot, and they’re being very polite, you know. He’s making eye contact, even though it’s really the equivalent of his hands free radio. And we’re seeing, and they back up, and they’re doing this, and it’s like people interacting. And he would look at the robot, and he would say something. He said, “Now, you hear me, right?” And then he would look at the robot as if they were two different things. And it was amazing. So Cliff Nast at Stanford said, “Yep. People do that,” computers as social actors. If it moves we think this part of the brain just says, “Yeah. It’s alive and it needs to behave, rules that living things do, otherwise it creeps us out.” We ran some more formal studies and, indeed, we started seeing that, and that eventually led to Cindy Bethel’s PhD. She’s at Mississippi State, that it does make a difference how you interact.
So here you’ve got-- we’d always thought, oh, you know, you would only react socially to a robot. If you did it’d only be if you made one of those creepy humanoid ones. No. We were showing that if you just drove up to a person with your bright lights, and you were in this sort of thing, and you violated their personal space, their blood pressure would rocket, their respiration would rocket up. They would get these grimaces, like “Get this out of my face.” But if you did the same move, got the same information from the doctors evaluating a trapped victim, or the same for the structural engineers who were looking around, but once you get within various zones, you slow down. You don’t violate the personal space. You change your lighting. You change the intensity of your volume, no problem. They loved it. There was no problem, the other way, so silly things like that.
You would think that would be, you know, who would have stumbled. We would have never come off of that, off a scientific reductionism, where we’re sitting around thinking about what are problems, and what’s the next big problem. It’s the ones that, because it’s so new, being in the field, having the problems literally leap up at us. We’re like, “What did we just see?” And we saw it again, you know, and to work on those. That’s what drives our research in my group.
Experiences in Real Disasters
Šabanović:
So how do you end up in the field, in a sense? Like are a lot of the things that you do, you know, part of exercises that people do? Do they call you up when something happened? I’m just curious, 'cause you have a very unique kind of expertise. I’m just curious.
Murphy:
I tell people that Disaster City is my favorite place on earth. Texas A&M is also the state agency for Urban Search and Rescue, and they have a huge training complex, of which one, just one part is called “Disaster City,” that’s built to train responders in building collapses, all types of building collapses, all types of anything with urban search and rescue, any manmade structures, how people do things. There are two different types of train derailments, chemical hazards, multi-story commercial buildings, regular houses with basements, all of these things there. So I’ve been using Disaster City for years, and more importantly it’s not just the props, but they have all these top responders that’s home to Texas Task Force 1, which is the state Search and Rescue Team, as well as the FEMA Texas Task Force 1, so they also have all these trainers. They train about eighty thousand people a year in urban search and rescue, and so they bring in trainers from all over the world. And so we get access to their expertise and their facilities. We also have been in seventeen disasters. It continues to mushroom. As groups hear about us, we bring them robots to use. We offer them advice on how to use things, and in return we get the data and learn more about this. We go to their exercises. We do a lot of quid pro quo. We’ll bring you robots. We’ll train you on them. We’ll give you awareness. We can even give you continuing education units.
You use them and let us watch, and see what’s working for you and what’s not. And then we’ve developed different field methods for getting that information. You know, some of the stuff that they can do at NASA, “Okay, stop. Now we’re gonna give you this test of comprehension.” Well, that doesn’t work, okay, for us. So when can you suit somebody up and get those biometrics? When do you get it by observation? When do you get it by video recording everything? I mean, you know, if it takes six months to two years to go through all the tapes, it’s not very useful, practical way of getting information, so we’ve really worked a lot on metrics and data collection.
Šabanović:
And can you tell us a little bit about some of these experiences that you’ve had in real disasters?
Murphy:
We’ve been in a variety of disasters. Of course, the first was the World Trade Center. That was amazing and, you know, a pivotal moment, because it showed that robots could be used and used effectively. They were being used not, you know, there was really no chance of survivors. But if there was a chance, it was gonna be in the basement or in the stairwells, and they had no idea where those were. I mean, you knew where the basement was, but you didn’t know what the collapse looked like. You didn’t know the shortest path. So given that, the responders, all the structural people, just have to take a guess of how it collapsed and say, “Well, the shortest path to the basement, probably with the least rubble to cut through and remove, is,” you know, here. You start that at that corner. You start at that corner. Bring your crane and start there. So if you could find a shortcut through that rubble, that would be hugely important. And they were finding signs of these box beams, which were about this big, and hollow, were acting like soda straws into it. So if you could drop down in there, maybe you could work your way in and get deeper, and that was what we were using the small robots for, and checking with that. So that was really a big one.
Hurricane Katrina was a big one. We had been part of Hurricane Charley, realized ground robots really don’t do much for you. It’s a geographically large, wide area to search. UAVs would have been really good, and this was 2004, 2005 timeframe, where they were just coming on, just being built for the military special operations, and I had been fortunate to serve on the Technical Advisory Board for the DARPA Micro Air Vehicle Program, so I knew all of this, knew these were up and coming, knew what was being done. In the hobbyist community we’d already reached out, and so we had recommended them to the State of Florida for the 2004 really bad season. That didn’t come together, but Katrina, we were out there as part of the state team, and we flew at Pearlington, Mississippi. Couldn’t get in. We were originally asked to go to New Orleans, but couldn’t get in there at that point. Everything got cut off and locked down, and really began to learn about that. So that was again a moment where you saw how unmanned, small unmanned aerial systems could be used. That was a big moment.
Lots of other disasters, I think probably the next big one was the Tohoku earthquake, and the Fukushima Daiichi nuclear accident. I was not physically at Fukushima. Texas A&M made it totally clear that if I got within fifty miles of it, both myself and my family’s, all of our medical insurance would go away forever, which was fine. I was up in Tokyo assisting with the teams that were going in using the Honeywell T-Hawk UAV to do radiological surveys in structural and recon. And so at that point, because of all our work with hurricanes, and our exercises, we’d had the most experience flying by urban structures and doing inspections with small UAVs, so Westinghouse brought us in. And we said, “Okay, guys, this is how it’s gonna be different than trying to track a car, you know, like in “Homeland,” where, you know, you’ve got a drone trying to track a car across the desert, and is it going this way and that way, very, very different style of flying very different way you look at it, and how you set up your team and crewing, what you’re gonna look for, where are the ways to expect things are gonna fail or cause stress? And so that was we were brought in, but seeing that. And then we came back with unmanned underwater vehicles to help with the tsunami, the port clearing, searching for, you know, so many people were swept out, looking for victim recovery. But all of this is taking place during the cherry blossoms, you know, this beautiful springtime. So you would go at Minimise [ph?] Rico [ph?]. You know, you’re kind of on a hillside, and so from here down total devastation, buildings partially collapsed, if anything’s left, rubble everywhere, cars flipped upside down, boats upside down. And then right above that, you could see the waterline. There’d just be this whole stand of cherry blossoms, just gorgeous in there, and that contrast between life and death, normalcy and destruction. You know, it just haunts our work and just makes us want to work, everybody on my team, and Roboticists Without Borders, the companies that work with us, just worked that much harder.
Collaborators
Šabanović:
You mentioned before also, kind of the beginning of your work in urban search and rescue, and disaster robotics, you mentioned Satoshi Tadokoro, and now again you’re kind of working with Japanese roboticists, so what are some of these, both kind of U.S. and international collaborations that you’ve been involved in over the time?
Murphy:
This is for IEEE, right?
Šabanović:
Yes.
Murphy:
Yeah, okay, so, very interesting. So after 1995, really the two people working it were Satoshi Tadokoro and myself on the United States side. And one of the strengths of the IEEE is that bringing together and enabling international collaboration. So Satoshi started working from the idea of let’s create a competition that will help encourage people to think about the problems there. I was working more on the how would you build the systems? What would you do, and began working more that side of it, and that progressed through. And then in 2001, when it was, like, the first use of robots, that became very clear that this was real. This wasn’t just, you know, oh, and robots, yeah, they can be used for search and rescue, 'cause a lot of times people will say, “Oh, yeah, what’s a good use of robots, search and rescue.” It’s like, “Oh,” because it sounds nice, right? But, no, this was real. This was real. And so Satoshi, myself and Paolo Fiorini in Europe, we founded the Technical Committee, the Robotics & Automation Society Technical Committee on Safety, Security and Rescue Robotics, that allowed us to start bringing together a small cadre of about a hundred and fifty people at first. Now we’re over five to six hundred people. We get attendance of a hundred, two hundred at our annual conference, and so that’s really a big deal to be able to enable this intellectual collaboration that’s independent. And the Japanese took a very different approach. They went for the we’ve got rubble. We want to build robots that are gonna lift and remove the rubble. In the United States we took the other approach and said the structural engineers want to see what’s underneath before. They take the pickup sticks. We have different density of populations and the types of buildings that collapse. We wouldn’t have the-- but who is there, you know, and what was gonna be the best use of technology? So there’s a great way. You see different attitudes and approaches toward it, you know, and so that’s been really great. No one group can do it all.
Šabanović:
And you also mentioned before a little bit about this question of perception, and how, you know, the human also needs to be in the loop. Could you tell us a little bit more of your work in that respect, because I know you’ve done work on things, like, what the operators also need to be able to-- what they can do, how they can interact with the robot.
Murphy:
Well, one of the first things we learn from field work, this also came out of the World Trade Center, but we had done some exercises with Florida Task Force 3 in collapsed buildings, and they continue to see, the robots, mechanically they’re kind of primitive, but they’re good enough. You could use ‘em. What we were seeing were the bigger barriers was that the operator could not figure out what they were looking at, you know. First off, you’re at the height of a squirrel. It’s like, you know, that’s not a very-- we don’t think in that viewpoint. We don’t think in that. You’re in a deconstructed environment, so even if it was something that a person could move around in, it’s so broken up it would be hard to understand what you’re looking at, anyway. You’re often, most of the time if you’re working in rubble, you’re actually looking down. You’re moving down. There’s a top entry, and you’re doing a lot of vertical work, and typically we don’t spend our life rappelling around our environment. We work this way, not that way, so it’s very confusing.
And so we began to realize that we were seeing that if we worked together in teams of two, kind of you think of this one to drive and one to look, which you might think if you’re going to a new place, you’re gonna slow down a little bit, and your passer’s gonna look for the sign or the address. And you both are gonna look, but this person’s gonna do more, and you’re gonna do more driving. If you did something like that, and we solidified it into a protocol called “Lover,” worked way better, worked, in fact, nine times better. Jenny Burke’s thesis showed that two heads are nine times better than one. So we started seeing that, and getting that perception, and if a person can’t see it, often times you can’t have the robot see it, and we were seeing a lot of techniques and were like, “Oh, the robot’s got to be fully autonomous.” And it’s like, yes, we’re gonna use a Kinect right here. We’re just gonna take this Kinect on this robot and we’re gonna go through a collapsed building at Disaster City. Oh, look, bright light, total darkness, bright light, kind of weird light, yeah, yeah. That’s not gonna work, you know. Seeing what the actual conditions were and the perception, so what do we have to do to make that sensing work and different sensors? Another project that we discovered that, I mean, it’s not like you discovered this. It was kind of new, but none of us in robotics had really thought through what happens when you hit a big chunk of glass. There’s still a piece of glass up there. Lasers go through it, unless it’s dirty, yea, so it doesn’t see that there’s something there. And sonars and other things will bounce off and don’t come back to you. So, again, you’re getting false readings all over these very complex environments. And office buildings, so if you’re trying to go through rubble in office buildings, you’ve got carpet. You’ve got acoustic ceiling tile. Let’s think of that concept, acoustic ceiling tile to dampen noise, so it’s absorbing some of these acoustic signatures, these things, carpet, things absorbing light. What a challenge, what a challenge. So, again, it pushes us back. The human is mediated by a computer, so they’re handicapped. The sensors are handicapped by the environment. What a great challenge to see who can be smarter faster, more effective. How can we build systems to make it where the robots can be much more effective?
Šabanović:
What would be some of your, kind of achievements, I guess I’d say, that you’re most proud of in working in these, both in the field work and in this more controlled lab work that you’ve done? What are the things that you feel are the?
Murphy:
Our work really goes into two areas. There’s the applied work, what we’ve learned that we can immediately feel. Our work goes into two areas. There’s work that, things that we learned that can be immediately applied to what the responders, the emergency workers are doing, and then there’s the more basic research. The applied research is just showing how it can be used, coming up with these protocols. Look, two heads are better than one. Two heads are nine times better than one. Do it this way.
Run a sterile cockpit. If you’re going to use a UAV, this kind is good for this. This kind is good for that. Coming up with these heuristics, if you’ve got a rubble looks like that, these robots won’t work. These are gonna be that, you know. There are metrics now.
We’ve been able to create metrics and really quantifiable methods of doing that. On the intellectual side, on the fundamental research, I feel like we’ve made fundamental contributions in human-robot interaction. We always thought about-- the field’s always thought about robots being people behind the robot, as if you’re doing field robotics.
Nobody thought about, well, if you have the robot interact with somebody in the field, how does that do? I mean, we’re thinking about that for entertainment robots and healthcare robots. This is a very different situation. So the fact that these aren’t anthropomorphic robots, they’re interacting in a very different way, so helping create that as a whole new theme within human-robot interaction. And some of the protocols that we’ve developed, that’s a fundamental contribution, of itself.
Šabanović:
You mentioned some of your collaborators already. You mentioned Cliff Nast, and Toshi. Are there any other, and some of your students; are there any other collaborators or students that, you know, that you worked with a lot, or that made an impact on your thinking and your work?
Murphy:
I have been so fortunate. You know, as the professor who’s the head of the lab, the head of the Center, you get all the credit, but it’s really the students who do all the work, and I’ve had phenomenal students, and I have phenomenal collaborators.
We’re seeing the first set of students that were with me at the World Trade Center. One of them is now a DARPA program manager, Mark Micire, who’s working on very small, fast, autonomous UAVs, which, by the way, are not only supposed to fly indoors, but indoors in damaged buildings, so, you know, very pleased to see that. Cindy Bethel’s PhD work, even though she does more social robots, came out of that. What would you do with a trapped victim? The answer was if you let a doctor drive the robot, or a roboticist, if we drive it the way we normally would, or program it to drive the way we normally would, it would be creepy. Yeah. It’d be creepy. And simple things, going, again, back to the cognitive and the social ethological literature, how do you do that?
We’re really excited about Britney Duncan’s work. She hasn’t graduated yet, as of this time, you know, is finishing up, but the idea was let’s use these small UAVs for indoor flights for evacuation. How would you zip around? You know, do you guide them? So first off, go talk to the evacuation people. They need less to, “Oh, follow me, follow me here.” They need more, “No. Don’t go this way.” I mean, that’s one of the big things you want to do, is you want to block off the ways that you don’t want them to go. So a robot could come way over here to that, and then like an angry bee or wasp, zzz zzz zzz, you know, kind of encourage people to go the other way. So how would we do that? And so she started her very first set of experiments with that. You know, she went back to the biology and looked at how wasps, and insects, and birds do things, and even mammals, how deer, you know, kind of protect their nest, and different things, all cool stuff. And so she was doing the first thing of, like, would you react to this, or would you react to?
People weren’t reacting to a quad rotor with open, spinning blades coming at their face. We know people get creeped out when a robot comes to fast on the ground. They weren’t showing the same pattern with UAVs, and they should. All the theories predict it. Cliff’s stuff predicted it. Why? But they didn’t, and what we think is, is that we’ve grown up as a species, no, no, no, you know. If that’s a lion, or a tiger, or a bear, you know, things that come after us on the ground, we know those are coming to eat us, or not, you know. It could be a nice doggy, but maybe not. But, you know, birds, yeah, pain in the ass. Insects, ech, ech, ech, you know, not enough to have a, “Oh, my whole survival depends on watching out for what that hawk’s gonna do.” Right? You know, so we’ve had to interact. Think about what that’s gonna do for consumer safety? We’re already hearing of these accidents, where people are getting too close. We’re already, you know, people, “Oh, we’re gonna do selfie robots,” you know. Yeah. That can work, but you’ve also got to take in the fact that people aren’t gonna get out of the way the way that they normally do. You can get some big screw ups with that. And so what exciting things that come out, again, totally not what we expected. We had a whole sheet, a whole set of experiments printed, and it was like, get back to the drawing board, and some things there. So those are some of the things that we get to do with the students, but there are so many of them, it’s just hard to. Josh Peschel has taken what he’s learned with UAVs and is applying it to water vehicles, building these incredibly small, little aerial, I mean, a air box, you know, airboats, to go back. He’s already got grants from the National Science Foundation, The Gates Foundation, how water is being used, irrigation. Yeah. You say you aren’t using it, except for the fact that we can now get back up in here and see you are, or that there’s pollution coming into these streams that you can’t even, with a kayak, get back in, but things this big can. What exciting times. All my students have done so well. We’re so excited.
Future Challenges
Šabanović:
What do you see as some of the future challenges for you?
Murphy:
You know, for me, what I still see as the future challenge is as the whole field begins to realize it’s a lot about unmanned systems, but it’s a system, there’s a part that’s the robot or robots, and there’s the part that’s us, that it’s all part of a human system meeting human needs. And so we have to design these systems to work together, or a system where all the parts come together. It’s no longer we no longer get points for making the best wheels, or the best flippers, or the highest resolution sensor. If it’s not usable, if it’s not useful for a particular application, if I can’t use it for responder, if it’s gonna take four months to learn how to use it. If it’s not reliable it’s not gonna be used. Those are the challenges now, as our field matures, and the fundamental question, how do you take something that you don’t know what it’s gonna be used for, and responders who’ve never seen, they are kind of nervous, how do you develop methods that allow both groups to quickly converge on good uses? Theory of diffusion and innovation says those processes can take twenty years. You can cut that down by half, even more, if you can get it in front of the users. They can adapt it. They can do things if you can go this coevolution round. How can we formalize those techniques for robotics, and how can we teach the next generation of researchers and R&D personnel, as well as our students, you know, who go directly into industry, to have these techniques of how to accelerate the adoption of innovation, and to generate innovations that just aren’t cute and interesting, but immediately a game changer. So that’s for me the big picture for the future.
Advice for Younger Generations
Šabanović:
And so one more question we have is actually about what you would suggest to students who are interested in robotics, or people who might be interested in pursuing, you know, an educational path or career, what do you think? What should they pay attention to? What should they think about?
Murphy:
There are so many paths in robotics. I would hate to dissuade people from their chosen path, but in the group, and in the style of work that I do, I recommend breadth over depth. So I was up at a university giving a keynote talk, and the students all came in. I got to have lunch with them. They were great. They were fascinating. They were, “Oh, I’m here because I can immediately, all I’m taking are robotics classes, either control, or AI, or programming, and this, and I don’t have to screw around with any other types of majors, you know, or that.” I’m thinking, “Oh, great.” And that’s good for them. My group, what I brought to the table, the ability to do ethnography, which came from having taken an anthropology class, ‘cause I was in a liberal arts school for a while before I transferred to Georgia Tech, and I had to take some social science classes.
Well, that’s skill allowed me to see certain things and start, “Holy crap. I need to write this down,” and, “Oh, by the way, yeah, there’s that little format they use in there.” I knew the terms. I knew what to look for. This was not a surprise. I could apply that. Likewise, Cindy, she had had her minor, undergraduate minor had been in cognitive science. She was bringing in-- wait a minute, the psychology have all of this literature. Wait, I’ve heard of this. I know this, you know, yeah, yeah, yeah. Here it is, knew what to look for. Field work, just general stuff, a lot of students who have done very well with me have been Eagle Scouts, being outdoors, at some point learning how to tie a knot, and, you know, who’s gonna move the-- I remember one student looked at all of our robotics gear when we went to do field work. He said, “Well, who’s gonna carry it out to the field?” I was like, “Us, yeah,” you know. And so having some of that, all of your life’s experiences come to play into this. That’s what I love about the artificial intelligence approach in the fieldwork. You’re bringing everything together, so think about breadth. Don’t think about, well, I can only take math courses for here the rest of my life, because that’s the only thing that’s valuable. You’ll be surprised. As you get older, the older I get, the more stuff that I’ve seen, the more that brings into the mix, makes it. So I recommend breadth. I also recommend if you’re-- well, I didn’t do robotics as an undergraduate. I didn’t take any of that. Well, neither did I, and so get in there and learn. What you’re learning is how to learn, and bringing your experiences and your judgment in there. And all this brings something a little different to the table, and that’s also part of the research mix is we’re not all the same. We see things differently. We have different experiences, and that’s what invigorates, and makes working in groups, and having whole fields. Everybody’s different, and they come up with different things. We all get to meet and see what the other person’s done, and take that part and add it to ours.
Women in Robotics
Šabanović:
Thank you. So we’re here for a PC meeting, but it’s composed of all female researchers. So I also wanted to ask you if you could maybe comment a little bit on the positon of women robotics, and your experience as a woman, in a field that’s traditionally perhaps been seen as the more male oriented domain.
Murphy:
You know, I think the view of robotics comes from the engineering, mechanical engineering tradition, where it’s guys doing eh, eh, eh, or, you know, building a humanoid robot that’s very heavily controls oriented. When I started in robotics, I started in artificial intelligence for robotics. Artificial intelligence has always had a parity with women is equally it’s me, women, as guys do artificial intelligence, and the same thing for AI for robotics. You have some very big names, Mima Tariq, Lynn Parker, you know, there are a lot of us. So that was no surprise. So the idea that having a program committee that’s all women, it’s not surprising to me. It may be more surprising to the more traditional control oriented, because they don’t see as many women. And I’m glad it’s an all women’s PC, just to kind of show to some people who may not, women who may be at a smaller school don’t see as much as many women in these kind of roles in general, because we just have a low enrollment of women. No. There’s nothing inherently, you know, guy about robotics, either controls or AI, and there’s always been a lot of us on the AI side. There’s a lot of us overall in this and to be part of that. And as a field roboticist, I’m one of the few field roboticists, and I’m a woman. It all works out. No. There’s no particular barrier and shouldn’t be. And it’s very fun to see the number of middle school girls who relate to that, and are saying, “Yeah, I like to do this stuff. I can do that.” But there are just so many options for women, and I think that having the program committee and the people on it who represent every type of robotics is a big deal, a great role model, and I’m glad we can just show it off, you know, in this big way, because sometimes it does get lost. And, you know, the women coming up through the ranks don’t see it.
Šabanović:
What about the extremes as the woman in the field with the robots in these disaster situations?
Murphy:
So as a field roboticist, being a woman in the field’s been kind of funny in a way. I will go to a disaster and I’ll literally be the only woman with, like, forty miners at the Crandall Canyon mine disaster. That was interesting. But the fire rescue teams have women. You know, they have women fire chief. Women are all over the place. So, you know, it’s changing constantly. Being a woman has actually helped. When we first started out the fire rescue teams we made so many mistakes. It’s stupid stuff. Didn’t do our homework, you know. Said, “Oh, look, this robot will be great for you.” And they just watch it fail. And I think they would have just shot. “Go away now,” you know, and never returned calls, or oops, you know. But they felt sorry for us. I think they related to us as, “Oh, look, these are,” because my group has always had a large number of women.
These are like our nieces. We’re really proud of them. They’re trying, and we need to just help them out, and so Chief Rodger said, “Why don’t you take some of the training classes with us? Why don’t you actually do this and see what we’re trying to do,” and, you know, that.” I’ve talked to guys, and they’ll return calls. They’ll actually do that, you know. So it helped us in the beginning, and then we began to get the expertise, and things, so that was that. But, you know, it is what it is. You use your skills. You use your relationships. You use your knowledge. You use all of your world experiences, and that’s one of the fascinating things about research for me, being a professor, being an intellectual entrepreneur. I get to call the shots. I get to try the things out, and I’ve been so fortunate that with the great students and the people I work with, and the teams we work with, that we’ve had some successes.
Šabanović:
Thank you. So is there anything? I think this is it for our general questions. Is there anything that we didn’t cover that you would like to have included?
Murphy:
No. I think you got everything. Hopefully I didn’t set back, you know, women in robotics at all.
Šabanović:
No. Send women intellectuals into grievance.
Murphy:
Well, let’s not do that. The idea is that, well, you know, basically the guys from this university, or guy assholes, you know, knowing it all, and we were like, oh, and they felt sorry for us, you know, versus that, you know. That’s a sad story, but it’s true, you know. I hope that doesn’t come across demeaning to women. It wasn’t like we were using our feminine wiles. It was just these guys were clearly very proud. They could relate, particularly to my women students, that these were like their nieces, that they wanted, you know, you get in there and you get the good jobs. You go do this. We see that a lot with some of the fire chiefs we work with. They want their kids to do better and their daughters, and particularly in Texas where there’s like whole hunting goods places dedicated to women hunters. They’re just like, “Yes, because it’s so nice with all the, you know, Reebok and all these companies realizing that women make up fifty-one percent of the population, because I’ve got thermal underwear built for a woman, and I’ve got all sorts of things that I didn’t have when I was growing up as a tom boy. Anyway, anything else?
Šabanović:
Thank you. No. That’s it.
