Oral-History:Lydia Kavraki
About Lydia Kavraki
Lydia Kavraki is the Noah Harding Professor of Computer Science at Rice University in Houston, Texas. Originally from Greece, Kavraki studied for her Bachelor’s Degree in Computer Science at the University of Crete, later moving to the United States to earn her Ph.D. in the field at Stanford. Kavraki’s research has contributed to various subsidies in the robotics field, from physical algorithms to biomedical informatics. At Rice University, Kavraki is also a professor of Bioengineering, Electrical and Computer Engineering, and Mechanical Engineering. She also has her own laboratory at the university, dedicated to research in Computational Robotics and Biomedicine. Kavraki is a recipient of the 2000 of the Grace Ann Hooper Award, 2001 winner of the IEEE Robotics and Automation Society Early Academic Career Award, 2020 winner of the IEEE Robotics and Automation Society Pioneer Award and the 2020 ACM-AAAI Allan Newell Award.
This Oral History takes us through the university education and professional career of Professor Lydia Kavraki. Kavraki offers her experience growing up and studying in Greece and moving to America to continue pursuing her education. She then speaks to her accomplishments in research and her career in academia at Rice University. Lydia also talks about her many awards and honors, and her efforts to bring women to the forefront of the Robotics field. Lastly, Kavraki touches upon her network of colleagues and the many great relationships she has formed throughout her career, and her thoughts on the future of the Computer Science and Robotics fields.
About the Interview
LYDIA KAVRAKI: An Interview Conducted by Selma Šabanović, IEEE History Center, 24 January 2015
Interview #803 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:
Lydia Kavraki, an oral history conducted in 2015 by Selma Šabanović, Indiana University, Bloomington Indiana, for Indiana University and the IEEE.
Interview
Interviewee: Lydia Kavraki
Interviewer: Selma Šabanović
Date: 24 January 2015
Place: College Station, TX
Early Life and College Education
Interviewer:
If we could start with just getting your name and where you were born and when?
Lydia Kavraki:
I am Lydia Kavraki and I was born in 1967 in Heraklion Crete in Greece. I actually was raised in Greece and I did all my undergraduate education in Greece before I moved to the United States for graduate studies.
Interviewer:
Where did you do your undergraduate and where did you study?
Lydia Kavraki:
I studied computer science in a brand new computer science department that was founded in Greece, in Crete actually, the place where I was born, in Heraklion. This was a department that had just started. I think I was the second year of graduates from that department, so everything was brand new. The topic of computer science was brand new. There were very few computer science departments at the time in Greece, and that was considered one of the best because it had lots of new people who had come, who had studied all over the world, and had come back to Crete and founded this department. I felt it was an opportunity and I felt very lucky actually to be in that department and to study in that department for my undergraduate degree.
Interviewer:
How did you get interested in computer science?
Lydia Kavraki:
Well, that is interesting <laughs>. In Greece the educational system is such that you have to choose your department where you want to be admitted while you are still in high school. There are entrance exams that are nationwide and everybody competes for the same exams, and then you rank the departments that you want to go to, and depending on your grades you get your first choice, or your second choice, or your twentieth choice. It is a very challenging task actually for an eighteen-year-old to do this kind of ranking, decide not only what they want to study but also where they want to go <clears throat> <audio skips> …my parents helped me, especially my father helped me <clears throat>. My father is an electrical engineer and he was always interested in new things that are developing, and he knew a few things about computer science. He did not have a computer at the time himself, but he knew that something new was starting, so he encouraged me to go there, together with my mother. My mother has an antique shop so she is not in engineering, but she is a very broadminded person and she wanted us to get a good education and this happened. It was a little bit chance that I chose that department and then that I had the grades to get in <laughs> that department.
Interviewer:
What kinds of things did you learn there? How did you get interested in going to the U.S.?
Lydia Kavraki:
Something that worked very well for me in that department is that because this department was getting organized at the time I was there the classes were offered and everything, but it was just two years in operation. It was still being organized while I was there. We got a lot of math and physics classes at the beginning and a few, let’s say computer science classes, because the labs were just not set up for the computer science classes and the hardware classes and all that, and I found out I was actually pretty good at the math and the physics classes, and I was not intimidated by a number of my colleagues, my fellow students at the time, that had a computer and that knew how to program, for example, and that knew how to do things that I had no idea because I had never set my hands on a computer, but I felt comfortable. It was a good start in some sense, and then when we got into the computer science classes everything seemed easy because we had done much more difficult classes before starting the computer science classes, so everything started easy. It was a good transition in some sense to computer science; things made sense and I found it very exciting. I liked it and for some reason, I always wanted to go to graduate school. I felt that clearly the end of my undergraduate studies was just the beginning of something. I never considered taking a job after finishing my undergraduate studies, so I applied widely. I got help from my professors. Actually, I got a lot of help from them, and they told me about places that I could go. I went through all these tests and it worked. Actually it worked very well because I ended up at Stanford.
Experience at Stanford
Interviewer:
When you got to Stanford what was the environment like? Who was there? What kinds of <inaudible>?
Lydia Kavraki:
I found everything wonderful <laughs> when I went there. I thought it was just wonderful. I remember that at the time I went to Stanford, we did not have the internet to just go into the webpages of the department and look at what is going on, and who exactly is there, and how the environment is like, and the area, and the university. But, yes, we had information, we had books, but we did not have the kind of experience that a student can have now when they are applying for graduate school. I found everything wonderful. I liked the place <laughs>. I liked my professors. It was a good year in some sense. Sixteen of us started that year and I felt very comfortable in that environment. I was also very well prepared for it, so I did not have difficulties with the classes. Of course there were many things that I did not know, but the ones that I had taken before I felt very well prepared for these subjects, and at that time everything was interesting for me. Now how I ended up in robotics is a different story. I was always fascinated by shape and motion, and I was looking at different topics when I was looking for an advisor at Stanford, and I tried different things, and at some point I ended up in the robotics lab, and I loved it. I loved the people there. I loved the topic. It was something close to what I wanted to do. I wanted to have some interaction with the physical world in what I would do. I did not want to have to work in a topic of computer science that is completely disconnected from the physical world. It made a lot of sense for me, and I was lucky because I was at a good place at the right time I think.
Interviewer:
Who else was there when you got there?
Lydia Kavraki:
Well, the first semester when I got there, one of the first classes I took was from Donald Knuth who was still teaching at the time and I was so scared going to that class. Every time I was going to this class I was well prepared and I was trying not to miss anything and he was just extremely nice, and when he learned I was Greek he had me repeat all the Greek symbols <laughs> to make sure that he had the pronunciation right. But there were a number of people there, the robotics lab at the time, Jean-Claude Latombe was there and Oussama Khatib, and they were driving forces for the lab, and then there were other people. For AI people you have Yoav Shoham who was very much involved in the lab at the time, vision people, but also theory people. I mean computational geometry people for example, like Leo Guibas and Rajeev Motwani later, all these people created an amazing environment that combined different disciplines within computer science, of course, which I liked very much, and gave different perspective to the problem that I wanted to solve.
Interviewer:
Who were some of the people that you ended up working with closely?
Lydia Kavraki:
I ended up working very closely with my advisor, Jean-Claude Latombe, with Leo Guibas, who is a computational geometer, with Rajeev Motwani, whom really gave us a very different perspective on looking at the kind of work that we developed. I ended up working with several of the students, the post-docs that were there, notably Danny Halperin, for example, who was visiting from Tel Aviv, and so many people. It is hard to say, to name all these people. I am sure I am going to forget <laughs> many of them even if I tried to.
Interviewer:
Who were some of the students in your cohort? You mentioned sixteen people came in at the same time. You do not have to mention all sixteen <laughs>.
Lydia Kavraki:
Yes. Actually I have not kept track with them. They have done wonderful things in their lives, and it was the same here as I came over with one of the other women was Daphne Koller, who is now a professor at Stanford, and it is hard to keep up with people after a while, but definitely there were people who are now professors or high up at Google and other companies, so it was exciting.
Interviewer:
What did you end up working on in terms of projects while you were at Stanford?
Lydia Kavraki:
I ended up working on the path-planning problem. This is a core problem in robotics, and it is a problem of taking a robot from an initial configuration to a final configuration without colliding with the obstacles and of course respecting all the constraints, the physical constraints of the system. Maybe I should put a little bit that in context in the following sentence. I feel I was very lucky I joined the robotics lab at the time that I joined, and the reason is that starting in 1981 or 1982, there were some new ideas that were developed for the motion planning problem. Tomas Lozano-Perez in 1981 came up with the idea of configuration space, which was a really an amazing idea. Now everybody refers to configuration space, we do not even reference him sometimes, but looking at the problem from that perspective I think opened new ways of thinking about it and solving the problem. That was why soon after that we had people who actually— Schwartz and Sharir came up with exact algorithms for solving the problem, but unfortunately they were very expensive, exponential in the degrees of freedom of the systems, something which means impractical to implement. And then John Canny, who was at Berkeley, around 1987 I think, published his thesis which is a landmark in the literature and he managed to give an amazing algorithm for solving the motion planning problem. I was given this thesis to read it and I read it and I thought, my God, this is hard; If I am asked to implement something like this I would never finish my PhD. But was a really brilliant piece of work, and then the lab I went into had kind of assimilated all these ideas and was trying to look at randomized techniques for solving the motion planning problem, and when I got there they actually had their first successes. They had a planner which was called a randomized path planner that was really the first planner that combined these ideas from configuration space with ideas from potential fields that were developed by Oussama Khatib and was successful in solving problems with robots that had more than three or four degrees of freedom in a reasonable amount of time. This was the problem I was given to investigate and at a very good time. The lab was enjoying a tremendous success. They were among the first to be able to solve very complex and what was considered at the time very complex motion planning problems, and I found it very intriguing to work on this problem, and I worked on this problem for many years developing a new technique, introducing yet another idea. It was the right time in some sense, the right place at the right time for developing these ideas, and when I look back at that time it was really the time was ripe for these ideas and there were other people in Europe and other parts of the world that had started thinking in that direction, had started thinking about using sampling to solve the motion planning problem because this is what we did. We used the sampling which seems totally crazy from one point of view, but it worked very well, and we developed these algorithms at the same time there was a group in Utrecht which was developing similar, very similar techniques and we actually collaborated with them, and then we had an approach to the motion planning problem and new ways of solving it that, not because of my work, but because what the community did. It really took off and gave us, now we have hundreds of these planners around <laughs>.
Interviewer:
What was the time period that you were in Stanford for?
Lydia Kavraki:
I was there from the 1990’s to 2006 or 2007. In 2007, I moved to Rice University and I was offered an assistant professor position at Rice University.
Interviewer:
Were there any other projects that you participated in in Stanford?
Lydia Kavraki:
Yes, there were, and the lab I was in it was a very collaborative environment and something I thoroughly enjoyed during my time there as a student and something that I tried to replicate myself now that I am a professor. I had my main project on motion planning, but I worked on assembly planning with people in the lab, Randy Wilson and Danny Halperin at that point, and my advisor, and Leo Guibas, so a number of people were involved. I also had the chance to work a little bit on a motion planning for stereotaxic radiosurgery which was another project that my advisor had started at that time with colleagues at the medical center, and this involves how to move a robot, radiate a brain tumor without really bombarding with radiation very sensitive tissue in the brain. These immediately come to mind, but there were many smaller projects I worked in and when I finished my thesis I started working also and applying motion planning techniques to proteins and molecules.
Interviewer:
What was your position after you finished your thesis?
Lydia Kavraki:
After I finished my thesis, I spent six to eight months as a research associate at Stanford, and this is when I started working on the application of robotics-inspired, robotics methods to drug design, to protein engineering, and it is a topic that I still continue until now, I liked that also very much.
Interviewer:
What did you like about it? What seemed interesting?
Lydia Kavraki:
About the application, this is your question? There were many parallels in these two problems. When we just think about it, when we have robots moving in a physical world and then these proteins, how do they connect? But there is an amazing connection. First of all, abstractly the problem is that you are trying to solve a geometric problem under physical constraints, with motion planning we have to respect the geometry, but we also have to respect the physical constraints, and this is what makes robotics very different from other parts of computer science, and this was actually what was very attractive for me working in robotics. If you are transporting a glass, I mean, a cup, you cannot let go because it will fall down. Gravity, friction, things did not work the way you expected <laughs> them to work. If you go down to the molecular world you are talking about, it is at a very different scale. Here we are talking about an angstrom, things that are measured in angstroms, and angstrom is ten to the minus ten meters, but definitely they have a geometric shape. They have flexibility which can be modeled in ways that we model the rotational joints in robots; they cannot be actuated. The molecules cannot be actuated, but they live in a world that they are influenced by energy fields. The part that connects with the physical world is again there, and it makes things much harder than they would have been otherwise in a very different form, and I found this interplay very interesting. When we started on that project and I started on that with my advisor, Jean-Claude Latombe and a very amazing person from Pfizer, Paul Finn, who was helping us understand what was going on at the time, and Michael Levitt was also around and he had a theory of force fields. It was just amazing to listen to Michael Levitt discuss how he folded proteins at that time and what he thought would happen. We thought that we would take our robotics methods, because we had been very successful with our robotics methods, and just apply them to small molecules that we model as robots and we will find whatever we need to find. We will find low energy conformation of the molecules. This was the first call, we will explore the conformation space which is analogous to the configuration space and we would be done in six months, <laughs> that was I think because we were so excited with the successes of the sampling-based methods that we introduced. We went from solving problems that had four degrees of freedom, five degrees of freedom, to problems that had ten to fifteen degrees of freedom in reasonable amounts of time, and rather consistently I would say. There was variation in the running time of our solutions, but it was not as bad as it was before the introduction of these methods. We thought that we would take the probabilistic roadmap planner, which was our baby planner at that point, and apply it to molecules and we would explore the conformational space of molecules, and then it turned out that that problem is an extremely difficult problem, and typically you do not have six or ten degrees of freedom, but you have a few tenths, and if you want to look at a small protein, you have few hundreds, and if you want to look at the reasonable system you have few thousands, and now that we are looking at viruses you have many <laughs> thousands, and it is an extremely complex problem. But I found it very intriguing to go back and forth between the two different problems, and I can tell you that this has been very inspirational for us. We had methods that we thought were really great and we applied them to another domain and they just did not work, and then we dug down deeper and we tried to understand why is it that they did not work, and we understand that we needed new ways of dealing with data storing, indexing it, and we developed methods that we took back to the robotics domain and then we had even faster path planners for robots, and that was very exciting.
Beginning work with Robotics
Interviewer:
What were some of these methods that you then translated back to robotics?
Lydia Kavraki:
We translated back to robotics different ways of searching locally the conformational space, that was one idea. Different ways of moving, but locally. Storing the data, that was another idea because the spaces have very high dimension. You need to have some idea where you have gone and where you have not gone, and it is very difficult to keep this information around. We ended up devising dimension reduction methods and we kept this information in a reduced dimension space and we found ways of going back and forth between the reduced dimension space and the original space, and this all translated to advances into new algorithms and new sampling-based algorithms for robotics, and in some cases it turned out actually that maybe because the reason we were successful is that many of the problems that we wanted to solve in robotics are not that hard. They are very difficult from many points of view, but they are not pathologically difficult. When you have a robot in a space you are going to expect this robot to do reasonable motions. You are not going to expect this robot to inspect, all the pipes in this building. You build a very specialized robot for that purpose. If you want a robot that will pick things up from this table, you do not expect this robot to be doing something very strange, and you do expect that the space would be engineered in a way that would allow the robot to actually access the table and to pick up things from the table. Very difficult problems, but not pathologically difficult, and exploring this and understanding this and exploiting this in the design of our algorithms was a major thing. I think it was a major accomplishment. Once we understood that, it was not us but the community replicated the techniques and found out that they worked. Then we started with a planner which was called the probabilistic roadmap planner, that was around probably 1996, but then later I remember my advisor saying at the time you have to make it simpler, and I was afraid of going to make it so simple that I am not going to get my degree, <laughs> but he was right, and it was critical that we simplified, we rid the robot of all the extra stuff, but yes, they could be very useful and they can increase performance, but they are not the crux of the matter and we presented the approach in a way that it can be explained in a minute, in a few minutes it can be explained to the students. I think this motivated many people to try the techniques because I can understand, I can implement it, I can do something with it, and people did things that we never expected they would do. Yes, they added back many of the things that we had deleted at some point because indeed they had performance for several cases and they were interesting ideas. The community developed methods that we never thought of and it took off. After I would say 1996 for the randomized planner, and then two years later we saw many people working on, not only on that technique, but on similar, techniques that had these new ideas. James Kuffner, who is now actually leading robotics at Google was at the same lab at the same time, and Steve LaValle was also a postdoc at the same lab at the same time and they came up with the idea of expanding a tree into the configuration space which was very fast. It was an amazing way of looking at the problem and I was just surprised. Honestly, I was very surprised to see so many people taking an interest in this problem and coming up with all sorts of ideas that were good and interesting and really made a difference, and at some point actually it got very confusing. I would say that about ten years later, 2006, 2007, 2008, I started thinking this is getting really confusing here. I cannot follow what is going on. There are so many ideas and there is so many variations in this domain we have to do something. I discussed it with my students at the time. We had so many people working on this topic. There were brilliant implementations out there and there were not so good implementations out there, and it was very difficult to distinguish what worked and what did not work because it is not only the idea, it is also the implementation of the idea and how easy it is to use it or to tune it sometimes, I think it started in 2006- 2007, one of my students, Ioan Sucan, went to Willow Garage and they started the idea of developing a library of sampling- based motion planners. Actually at the lab, we were in the fifth edition for library for sampling-based motion planning, but nothing was good enough. Everybody would write things from the beginning and would try to implement it in a better way, but Ioan took possession of this project and Willow Garage helped, and Mark Moll, a longtime collaborator of mine, got involved in this project, and really it is them who developed the Open Motion Planning Library, which is a plan, it is an open source library right now. It implements sampling-based motion planners, not only the planners that we have developed in our group, but planners that had been developed by the robotics community. It is, as I said, an open sourced library so it was very nice to see at some point different groups contributing to this project and implementing their planners. We tried to make it as much of a community effort as possible because this is the only way to sustain the project. We want this to be a community effort to implement these kinds of planners, the sampling- based motion planners. Mark and Ioan did an amazing work with the library. We had other advances at the time. ROS came in for the robot operating system and we connected it with ROS, so suddenly this library had been tested on thirty plus robots that we could not get our hands on, and now we are working with NASA and we want to import part of this library but also extend it in different ways for the Robonaut, which is the robot that NASA has actually got right now in the Space Station which is amazing and very challenging for us to work with robot.
Interviewer:
What other kinds of robots did you work with?
Lydia Kavraki:
Mainly we worked with manipulators because the techniques that we had developed made more sense for this kind of robot. They can be applied to any kind of robot, but it is an overkill for many applications and I do not see why one should be applying sampling-based motion planner to certain robotic platforms. For manipulators it makes sense and it has been very interesting to see how things developed. Initially, we understood very quickly that we cannot apply this approach without some kind of smoothing. We understood later that we have to take into account other things, especially now that we are talking about cooperation of robots with humans. You have to build many more things in your motion planner, even if it is a manipulator arm that is working next to you in order to be able to feel comfortable and work actually with-- collaborate with a robot in a certain way. So mainly we have been focused on arms.
Interviewer:
Have you done any work with the actual engineering of the molecular structures?
Lydia Kavraki:
Alright, let me move into that. Since I finished my PhD, I have been working on this topic. Initially, it was looking at very small molecules and looking at drug-like molecules; drugs are typically very small. They need to penetrate all sorts of membranes in our body. It is easier if they are small molecules, and we worked with a pharmaceutical company. They did not tell us really what they did with the methods that we provided. Unfortunately, it is very difficult to understand how these methods had been used. The drug development process is an extremely complex process, it takes about ten years and about one billion dollars to develop a drug. The interest of these methods, the phase of the process where we can interfere is when you select candidate drugs. You can never, with computational techniques, select the drug that will work. This is very difficult, but what you hope you will do is that you will find from potentially millions of compounds you will find the hundred or the thousand that have a potential to make it down the pipeline. We looked at small molecules, understanding how flexible they are, then we looked at docking the small molecules into receptors. The way a pharmaceutical drug works is that typical there is a protein. There is a cavity in that protein that helps a reaction to take place, and if you can engineer a small molecule that would sit tightly, will fit in there and will sit tightly in that cavity, then you stop the activity that is taking place in that cavity. A famous example is the HIV-1 protease which is a molecule that has two flaps that open and close. When the virus affects a cell it uses the cell machinery to replicate its own genetic material and then the material of the virus is being produced, this protease is cut into pieces, and when they cut it new copies of the virus can assemble, mature, and infect more cells. When you jump these mechanisms, it is like scissors, similar like scissors. If you jam it, it does not work, and this is what drugs try to do, they try to jam these mechanisms, and we used geometric reasoning coupled with energetic reasoning to develop methods that would dock small molecules into cavity of larger molecules. Initially it was small molecules, but I can tell you that now very recently in our work we are looking at fairly large peptides for drugs because the knowledge has advanced and because maybe some of the easy questions have been solved and now people are looking at more complex diseases and actually diseases like, for example, asthma, where longer peptides are needed in order to deal with the kind of receptors that need to be blocked. This was one aspect of what we did. We never did the bench work; the lab work was always done with collaborators. I am very fortunate that at Rice we are in a very interesting milieu, in the sense that we have two medical schools next to us, and MD Anderson Cancer Center, and the medical part of UT, and UT Galveston, which is a medical school. There is a huge number of people around us that deal with not only drug design, but understanding more or less the molecular basis of diseases, and we have worked on various aspects of the problem. We have worked on functional annotation, trying to understand if a protein exhibits a motif that is related to a function, a known function, it can be inhibited with certain drugs, and we worked on looking at very large structures, and these are typically viruses, pictures that are not obtained with traditional techniques like an MR or X-ray crystallography, but with cryoelectron microscopy. You freeze the virus and you take slices and you reconstruct this mechanism which is just an amazing mechanism. It may have two hundred proteins that form a shell and somehow these proteins collaborate and a shell can expand from a small size to a very large size to accommodate DNA material, and they infect cells and we have pictures of all that and we have no idea how they work. Not providing a definite answer, but providing a hint of how this mechanism works, can be extremely helpful for jamming the mechanism. This is what we have been trying to do and we treated it as a robot with very, very large number of degrees of freedom. We explored symmetry, we tried to simplify things, but to simplify in a meaningful way, not lose the important detail but lose a lot of detail because we just cannot deal with the systems, and we do apply motion planning techniques to these problems, and we have also worked with metabolic engineering. There has been a plethora of problems that we were lucky to encounter and work on, and this was always done with collaborators from mainly the Texas Medical Center.
Interviewer:
Can you tell us who some of them were?
Lydia Kavraki:
Yes, one who goes to Rice, George Phillips, who is a biochemist, was very inspirational to me and he presented a number of problems. Wah Chiu who runs a cryoelectron microscopy lab and he has shown us things that we never thought would be able to see with this molecule. George Bennett is a common collaborator, Cecilia Clementi. There have been just so many informal collaborations and discussions because of the density of people that exist there and because of the organizational structure that we have. We have a center called the Keck Center for Quantitative Biosciences, which is not a center with building, but it is a virtual center that brings people from engineering and people from medicine, natural sciences, to talk about problems that are multidisciplinary.
Move to Rice University
Interviewer:
Is this one of the reasons why you decided to go to Rice?
Lydia Kavraki:
Yes, that was very much one of the reasons why I decided to go to Rice. I like the department there, but I also saw a tremendous opportunity. I thought that I could continue my robotics work which I love and very much interested in continuing and that I could also have this other dimension in my work in collaboration with people from Rice itself. As I told you, George Phillips was one of the main people that recruited me there and outside Rice, in our close environment with the Baylor College of Medicine or MD Anderson.
Interviewer:
Can you tell us a little bit about going to Rice and are there other people that do robotics there? How did you start your lab?
Lydia Kavraki:
This was a little bit difficult for me at the beginning, because I left from an environment where there was a very well-organized and wonderfully working robotics lab, and I went to a place where I was the only robotics person and that felt very strange, <laughs>. That felt very strange at the time at the beginning, but it was interesting. We started the lab, recruiting students is always an issue when you start something new because you do not get the students who are interested in robotics to apply to Rice. They typically would go into one of the more established labs. This has taken time and it has taken a long time to establish the lab and to recruit people. Fortunately, the undergraduates at Rice were just so happy <laughs> that there was a robotics lab around and thrilled to have this opportunity to learn about robotics and they have been the driving force for much of our research. We have a number of papers with undergraduate students who have actually gone to conferences and presented the work, and sometimes I tell my colleagues if they are too harsh on the questions, sometimes I tell them, “This is an undergraduate,” and they go, “What? I thought he was finishing his thesis,” or she was finishing her thesis. They immediately are surprised by it but it has occurred multiple times so far. Later, Marcia O’Malley came in mechanical engineering and she does haptics, so that was very nice and James McLurkin, who does swarm robotics, was hired more recently. Electrical Engineering also has control people. Unfortunately, one of them left after I joined Rice. It has been a very small group, but I do not think this has prevented us from doing very interesting things. NASA is close by and they have very interesting problems. They had very interesting and very hard problems, and then the world is becoming smaller. Collaborations with other labs has definitely helped us move along, and for us, as I told you before, the collaborations with the medical center, so we have been very busy.
Interviewer:
Can you tell us a little bit about your collaborations with NASA? How did those get started? Who were you working with there?
Lydia Kavraki:
It has been very interesting. For the longest time when I went there, there were a number of discussions, it was very exciting, and then NASA was silent and we could not figure out what was going on. They were developing the Robonaut. When they were developing the Robonaut, they were not inviting everybody to see what they were developing at the time, and that was a strange period because we did not really know exactly what was going on, but after this robot has been developed suddenly we have an interest in path-planning, because now we have the robot, but what are we going to do with it? We want it to do interesting things, be as autonomous as possible, and it is a very challenging system to work with. Now we work mainly with the software division of the Robonaut project, and Julia Badger has been wonderful and Rob Ambrose in helping this getting started and nourishing this relationship over the years.
Interviewer:
Who were some of your other students, and you mentioned collaborators more internationally?
Lydia Kavraki:
While I was a student at Stanford, I visited Mark Overmars and his group at Utrecht University and they had very similar ideas, very similar planner at the time, and that is why we collaborated and we wrote together a paper that has been cited a lot since then. Peter Svestka and Mark Overmars were there and I enjoyed my interaction with them very much. Over the years, I have had contacts with the researchers in France. France has been at the forefront of robotics in Europe, and then Halperin, who I met originally at Stanford, moved to Tel Aviv, and again he has been a person that I have collaborated with. I have also tried to keep my connections and collaborate with people in Greece so I collaborated with Dr. Orphanoudakis and Dr. Argyros who were or are at the University of Crete.
Interviewer:
Who did you collaborate with in France?
Lydia Kavraki:
I had the opportunity to visit the LAAS and the group of Jean-Paul Laumond and Thierry Simeon and a number of people there. Florent Lamiraux was at my lab for a period of time as a postdoc and he went back as a researcher at last, and now again I have another postdoc from LAAS, from a younger researcher there, Cortés, and he joined two months ago, so I am very excited about this.
Interviewer:
Who have been some of your students who have also continued in robotics?
Lydia Kavraki:
Many people have come through our lab, and this has been just wonderful. From the graduates of the lab, Kostas Bekris is now at Rutgers University; he is an assistant professor and he is continuing working with motion planning, and many other problems. And Erion Plaku, who did some wonderful work while he was with us at Rice University in collaboration actually with another colleague of mine which has been a wonderful collaboration, Moshe Vardi, is at the Catholic University. There are two students who worked on the protein side Brian Chen, who is now at Lehigh, and Amada Shehu, who is now at George Mason University, and it is just wonderful to see all these people doing well, and Ioan Sucan is at Google right now. Andrew Ladd who was an amazing student of mine who gave us a new perspective of looking at the motion planning problem. Unfortunately, we lost him to cancer and that was a very difficult thing to deal with, personally and for everybody in the lab. It was very sad to see such a brilliant person go. We have people from pharmaceutical companies like Boehringer, to Google, and Amazon, and then postdocs. Many postdocs came through the lab. Many of them are now professors. They came as postdocs, but for example, Florent Lamiraux at LAAS, he is a researcher, full time researcher, and Oliver Brock is at the UT Berlin, and he is doing superbly. Nurit Haspel is at University of Massachusetts at Boston. I am afraid I am going to forget <laughs> them all. It has been wonderful to work with all these people over the years. This has been one of the most amazing things. To see them excited about the problem. To see them doing things that were, how do they manage to do these, and to continue now on their own with different ideas, different techniques, and new problems, different problems and the problems we have worked on when we were together.
Interviewer:
Can you tell us a little bit about your funding sources? Where have you gotten funding over the years?
Lydia Kavraki:
Funding has mainly come from National Science Foundation and the National Institutes of Health for our work in biology. We have had very little funding from companies. Willow Garage has also supported us, but relatively little funding from companies. But mainly from Federal Agencies.
Interviewer:
Are there any other projects that you would want to talk about that we did not cover?
Lydia Kavraki:
I think we have covered a lot.
Interviewer:
Just want to make sure. One thing that I was curious about, you mentioned early on while we were talking that there are kind of new problems now that robots are supposed to be interacting with people more closely, and I mentioned that maybe some of that comes across in your work with Robonaut. Can you tell us a little more?
Lydia Kavraki:
We hope it will come across. I think you need to take into account more constraints, but you need to understand what are these constraints that you need to take into account when you plan for a robot that is close to a human, and here we are not doing that, but human robot interaction people have a very important role to play to give us these data that we suspect it is this or that, but we need a more quantitative analysis in order to be able to have some data that we can use. I think it is going to come and it is a welcome development for a number of applications.
Progress for Women in Robotics
Interviewer:
I also noticed that you received the ACM Grace Hopper Award in 2000. Can you tell us about that?
Lydia Kavraki:
Yes, this award was a very big surprise for me. I never expected it, but it was awarded actually for the development of the probabilistic roadmap planner. This is the planner that I told you was developed at Stanford around 1994, we started having the original ideas. A major paper was published in 1996, and then the work continued very intensively for at least two more years, and this planner gave a new way of looking at the problem that was exploited by many researchers and really the award was not for me, but for the community I think, because it is the community that took this idea and expanded it in ways that we did not imagine at the time. This award was specific for development of that planner, and it helped us a lot. It was a wonderful recognition. I am very thankful for everybody who contributed and helped me receiving this award.
Interviewer:
Have you been part of the Grace Hopper <inaudible>?
Lydia Kavraki:
Yes, I love Grace Hopper. I remember when I was at Stanford, the sisters group was being formed and I remember at that time it was just a mailing list and when you joined, I was told by my friends you should join, and you actually sent an introduction with your name and what you work on. Now it is impossible <laughs>. I mean this is a huge mailing list. It is thousands of people; I think there were about eight thousand people last time. We try to send our students every year because it is very inspiring for them. I had the honor to give a plenary talk at Grace Hopper, and I have given many plenary talks, and I have given talks in front of huge audiences, but I can tell you when I got up to give that particular talk it was a different feeling because it was a sea of women in front of me. I had never seen so many women in computer science all at the same time. I had arrived the previous night and I did not really understand what was going on. My talk was scheduled in the morning and I went into the room and there were all these women there. I could not believe it. It was a very different feeling for the speaker <laughs> as much as it is for the audience I think.
Interviewer:
You are also part of the ICRA OC that is composed all of women and this has led us to also ask a question about how do you see the position of women in robotics and how do you see maybe increasing numbers or what are some of the challenges? What are some of the things that can be done?
Lydia Kavraki:
I think it is wonderful. We have been working well together. It is really Nancy Amato and Lynne Parker who conceived this and who executed it in the most wonderful way. We are very happy to help them. I think women have a pivotal role to play in robotics because robotics is a field that can be seen from so many different angles, and also robots in our society, what is it that we want them to do and what is it that we do not want them to do. Anything that has to do with interaction, but is important to have different perspectives and women bring a different perspective, as will different other groups. It is just important to have this diversity, and I think we work on some of the most interesting problems in robotics.
Interviewer:
As women?
Lydia Kavraki:
Yes. I think women are working on some of the more interesting and critical problems in robotics.
Interviewer:
Can you give some examples do you think? Not to put you on the spot.
Lydia Kavraki:
I would prefer not to go into details, because I am going to miss people.
Interviewer:
I also noticed that you were quite prominent in RSS. Can you tell us a little bit about that?
Lydia Kavraki:
Yes, there were two conferences that I have always enjoyed participating into. One is WAFR, the Workshop on Algorithm Foundation of Robotics, which started around 1994, and it covers many issues related to motion planning, which I love as a topic. I have enjoyed my involvement with that workshop and I have organized it in the past, and then around 2004 I was at one of the people who contributed to starting the RSS, the Robotics Science System Conference, I remember I was asked to chair the first one, but I had just had twins and I thought that is impossible. I was very lucky that ten years later they asked me again to chair the conference and this was last year. We had an amazing participation; a single session conference had about seven hundred people. We did not expect that. Pieter Abbeel built all the organizational work at Berkeley. It was good that he had reserved the big room, the biggest auditorium on campus, otherwise we would have had a major problem trying to accommodate people. We expected it would be popular, but we did not expect it would be that popular, and I enjoyed my participation also with RSS. It is a way to get to know the community, to get to the people in the community. One of the things I have enjoyed very much with my involvement in robotics are the people. I think robotics people are special <laughs>. I cannot describe it but I have always enjoyed working with people in robotics.
Interviewer:
Since you were there around the time of the founding of RSS, what were some of the reasons for the founding of the conference and who were some of the people who were involved?
Lydia Kavraki:
Sebastian Thrun was the driving force for RSS, and I think that he pushed a lot with Stefan Schaal, but it is hard to start enumerating people, but I remember Gaurav Sukhatme was there and people who wanted to see more things happening in the community and the community getting together and trying to look at what is happening with the field. I have a broad field of the field, but what is different for a conference like RSS compared to ICRA, for example, it tries to give you an overview of what has happened in the field of robotics during the last year, and it is very difficult to do that because there are so many different things going on.
Interviewer:
So you were mentioning about RSS.
Lydia Kavraki:
Yes, RSS is a very different conference than the ICRA. RSS tries to cover many different topics in robotics, actually tries to do another view if possible of the research work that is done in the field during the last year, and that is an almost impossible goal. I think it is very interesting to go to RSS because you get to see not only what happens in the field, the sub field of robotics that you are working on, but also the best of what happens in different fields in a short amount of time. You also get that through ICRA, but it is much bigger as a conference and more difficult sometimes to follow all the things that you would like to follow. I was happy to be involved and I got to know a lot of people in the robotics community. It was a good experience, a very good experience.
Involvement in Professional Societies
Interviewer:
Can you tell us a little bit about any involvement you have had with IEEE or Robotics and Automation Society?
Lydia Kavraki:
Yes, it is actually going to increase my involvement because I have been elected to the ADCOM starting in 2015; I look forward to working with the ADCOM and I have participated in the conferences for many years. I try not to miss the ICRA conference every year and the IROS conference every year, and if I can go to more I will, but we have constraints with our time, and also our travel budget. I have enjoyed the conferences over the year and enjoyed my involvement with people. I had the opportunity to meet a number of people. Oussama, he has been so wonderful introducing me to people and getting me involved in this committee and that committee and the other committee and in this book and in the Encyclopedia of Robotics that he and Bruno Siciliano have been doing which I think is a monumental project and really something very worthwhile for the field. I have enjoyed meeting the people and have been involved over the years in selection of awards, selection of organizational sessions, discussions, panels, all sort of things.
Interviewer:
You have also written a textbook on the principles of robot motion.
Lydia Kavraki:
Yes, actually I did not write the textbook, Howie Choset <laughs> wrote the text book together and he was a driving force. He assembled a wonderful team of people who helped. Actually, I did a very small part in this book. I did only the chapter on motion planning, and I described mainly the sampling-based planners, and if it were not for Howie <laughs> this book would not be there.
Interviewer:
We have a question that we ask everybody towards the end, but if you were talking to some students or young people who are interested in robotics or a career in robotics or studying robotics, what would you tell them? What are some things they should think about or pay attention to?
Lydia Kavraki:
I think it is an exciting field. It combines a lot of different topics, and I have thoroughly enjoyed this. You can work on the theoretical part of robotics, you can work on the mechanical part, and everything in between. You can work on robotics abstractly, or you could work on robotics hands-on. Also, with the way things are going, robotics and robots are bound to play very important role in our society, and I think we need to get more people involved. We need to get people with broad education involved and people who understand also the societal implications of what is happening right now and can help us transition to a better state. If you think about robotics in medicine, if you think about robotics for helping people, it is really an area that has enormous potential and potential also for bettering our society and we have a responsibility I think as a community to do this and to attract the right students and to educate the students to push the field forward.
