Articles

---

New tools and computational methods lead to collaborative, interdisciplinary solutions.

S.S. Iyengar is chair of the Louisiana State University computer science department, chair of LSU's Robotics Research Lab, is an ACM Fellow, and holds many other distinctions and received many other awards. In his appointment as an ACM Fellow he was cited especially for his research and contributions in algorithms and data structures in parallel computing, image processing and sensor fusion.

UBIQUITY: Tell us a little about your background.

IYENGAR: I was born in Bangalore, India, and came to this country in 1970 to Mississippi State University. I did my Ph.D. there in engineering. After that I had a passion to do scientific computing. My background is in mechanical engineering and so my initial interest was in mathematical modeling/biological systems. That got me interested in algorithms and data structure representations. I started looking into modeling and simulation problems. Initially, I taught at Jackson State University and then in December 1979 I joined Louisiana State University. This is where my career blossomed.

UBIQUITY: Two decades ago. How has your work evolved over that time?

IYENGAR: I started as an assistant professor with an interest in developing efficient data structures for image processing applications. One of the things researchers in late 70s were focusing on was efficient representation of images (image data structure) and potential image compression techniques. In 1980 I focused on this problem with Dr. Les Jones and we derived a very beautiful representation called Virtual Quadtrees. In addition to this we focused on some other data structure problems for balancing search trees. In 1983 I became an associate professor and my interest in computational science was changing towards application. In 1985 I was a visiting scientist at Oak-Ridge National Laboratory, where I started focusing on computational aspects of robots and more specifically we looked at the problem of path planning algorithms in unstructured environments. We had some very interesting results published in first-rated journals. Oak-Ridge National Laboratory supported my research for about 10 years. Here I can say that I have crossed the traditional boundaries of computing and this is the paradigm shift that I will be talking to you in this interview. Increasingly, applications require computers to interface with the real world and draw data directly from it. These applications range from defense to medicine, manufacturing to environment health. They all depend on inputs that are noisy, incomplete and of limited accuracy. So this paradigm shift in computation motivated me to look into multi-sensor fusion, which has emerged as the method of choice for implementing robust systems that can handle imperfect inputs. I coauthored a book titled "Multi-Sensor Fusion" published with Richard Brooks who is at Penn State University covering all the technologies and methods associated with multi-sensor fusion including multidimensional data structures, techniques for reasoning with uncertainty, approaches to enhancing system dependability, and working with meta-heuristics. So the period between 1999-2000 reflects 10 years of fusion research funded by Office of Naval Research (ONR), which made me introduce novel solutions to challenges such as image registration, distributed agreement and sensor selection. During these years I worked with scientists at the Naval Research Laboratory, the Jet Propulsion Laboratory, and many other places. I have had many Ph.D. students who worked with me on these projects and then went on to work at Los Alamos National Lab, the Jet Propulsion Lab, Oak Ridge and other distinguished universities.

UBIQUITY: Where do you see the computer science field going?

IYENGAR: The discipline of computer science will provide ubiquitous access to super computing, mass storage and regionalization through high-speed networking. That's a paradigm shift. So if you look at all of my areas of research and work there has always been collaboration with other people. It also has been multi-disciplinary and globally distributed. That made me look at research and education in a large variety of areas. I think that has helped me to look at a problem in a much more abstract level than just looking at computational perspective only.

UBIQUITY: Describe that paradigm shift you mention.

IYENGAR: Intelligent processing, a rapidly emerging discipline in science and engineering, aids researchers in developing systems that "see" and "comprehend" their environment using new computational methods and tools. So, the idea here is in the computation processing power shifts from local computing platforms to supercomputing centers. One example is the approach we use for doing a billion atom simulation process for material science research. This type of computation is possible now and we can do it at LSU using computational facilities available at Pittsburgh or Los Alamos, etc. So this type of technological change is certainly a paradigm shift in our computational process.

UBIQUITY: Give an example of how you use access to supercomputing.

IYENGAR: Currently, detection and tracking systems use a large number of different types of sensors. There has been an increase in the development of distributed sensor networks (DSN) for the process of information gathering. Availability or access to the supercomputing centers enables researchers in different areas of the world to collect and process information very accurately. Let's take a specific example of sensor fusion. The search for computationally efficient architectures that are suitable for distributed sensor networks is quite critical in the context of distributed computing because distributed image reconstruction procedures for developing multiple source locations as an energy intensity map depends on work carried out in several super computing centers. So this is again a classic example of merging different computational structures developed at different centers. That's again the paradigm shift. All of these things are done through high-speed networking, and so of course networking and communications are essentially the focus in many of these paradigm shifts. That means even if people are working even in very remote corners of the world on a problem of interest, you can still collaborate with them in terms of research or in terms of education. That's the kind of paradigm shift I'm talking about.

UBIQUITY: Put a rough date on this paradigm shift. When do you think it really started?

IYENGAR: I think it started around 1986 or '87. I can give you my own personal experience because I was at that time collaborating with some professors at Purdue. We never really met each other, but at the same time we were able to collaborate in the form of e-mails and things of that sort. I think we've been able to use technology to create these paradigm shifts. I proposed a very interesting data structure to a professor in University of Texas - Austin, David Scott. He and his colleagues immediately said "Yeah, it looks good" and then I said "Well. then why don't we collaborate on this?" We interacted intensively through e-mail, and that helped to formulate the problems and even provide solutions. Later on, he visited me at LSU. In my personal research this collaboration helped me to see problems in a much different way. As a matter of fact, I also wrote a paper somewhere around '87 with a professor Dekel working for IBM in Israel whom I never met. All of our correspondence was through email. He was at University of Texas - Dallas and then moved back to Israel. So we came up with a good paper on balancing search trees. I had been working on it alone, but he said, "I have an interest in finding a generalized solution to this problem." So that helped me to provide a global solution to a problem. I will always say that the paradigm shift helped me. I'm sure it has helped lots of other people in different aspects of computer science.

UBIQUITY: Of course, paradigm shifts are sometimes not recognized immediately, and even when recognized are not always understood. Do you think that this shift is now understood by everyone?

IYENGAR: I think it may not have been understood completely but I assume people are looking on this as a great opportunity to do lot of good working in different areas of computer science. But I would say the majority of the people who are able to use the enabling technology for this new paradigm are in high-performance computing and communications. For example, consider the relationship between the software engineers in India and the rest of the world. Engineers in India can now globally provide solutions to many complex problems. And naturally this paradigm shift helps to reduce the cost of developing software, because now you can always go outside and get something as good or even better quality at a cheaper price. That has opened ways for other developing countries to act in a very competitive way. So the paradigm shift is in pretty full swing. If I want to teach a course in networking to a group of students in a small village, they may not have access to the professors and so forth -- but I think now with this technology and the paradigm shift we can certainly do that in a very efficient way. At any moment, we can have a one-to-one interaction with the students too through distance learning. I would say the shift is only up to 80 to 90 percent. In the next several years, it will reach its peak. Everybody will have an access to this paradigm shift in computing, whether rich or poor, and regardless of location. In principle anybody at any place will have access to some computing.

UBIQUITY: In seven or eight years? That's an amazing prediction.

IYENGAR: It is going to happen for two reasons. I'll give you a simple example. I was in India last Christmas. I went to a village and saw a lot of public boards with the caption: "Learn how to write programs in C language." I could never have imagined this even four or five years back. These are not villages close to big cities where there is a lot of high-tech development. I'm talking about a very remote corner in the eastern part of India. When I see these public boards, that means that there is awareness and at the same time there are solutions. At least I can say in India, about 20 to 30 percent of people have access to Internet. In a few years it will be 60 to 70 percent.

UBIQUITY: Are you talking about 60 to 70 percent coverage in India only? What about other countries?

IYENGAR: Not only India. I'm talking about the Philippines and other parts of the world. Of course, France and Germany are pretty well advanced, but so are many of the Latin American countries I've seen. For example, I visited Peru couple of years back to give a talk and saw tremendous opportunities to do global communication. I'm really optimistic on this subject after having seen many of these places. This collaboration and multi-disciplinary, global distribution can revolutionize both research and education, and science and engineering, and even arts and humanities. In the United States every state in the union has special programs on information technology. I see even in high schools and middle schools that people are aware of it and using it. If this trend continues, if money continues to be allocated for it, I would say 10 years is a good timeframe.

UBIQUITY: Your optimism is interesting. There are certainly many pessimists who talk in gloom-and-doom phrases about the digital divide. You seem to believe that that's a problem that will work itself out.

IYENGAR: Well, there are always some impediments. Twenty years ago we never thought the Internet would revolutionize the whole world. People don't know what will happen. They are not that optimistic. For example, there is this area of what is called information grid computing. I see this type of computing percolating in every laboratory and every university. The digital divide perhaps can be an obstruction or maybe an impediment, but I think certainly that it can be rectified as we go along. I've seen in my research and my personal experiences in developing software that there is a lot of potential to be optimistic. That is my opinion and also the opinion of some distinguished scientists from different parts of the country and the world that I have talked to. I think a lot of good things are happening now. Of course, security has become an issue right now. And while there are no solid solutions, I think eventually there will be secured networking down the road.

UBIQUITY: In addition to security is there any other large-scale problem that you see as urgently needing a solution?

IYENGAR: In Internet computing, we need a way to give a nice semantic description of the search process because it is still not very user-friendly. I'm talking about having one hundred percent audience in the process, that means every person in this world should be able to understand irrespective of education or whatever they have. I think that is still not happening because if you want to search something on the Internet it takes an enormous amount of time because the search engines are not fast enough. There must be some hierarchy of structure that could be developed to make those things faster.

UBIQUITY: A Google search on your name pulls up a lot of information instantaneously. Are you thinking of other kinds of searches?

IYENGAR: Well, the names of people who have published papers or who are in the business will always appear. And there are not many people with my same last name. But suppose you were searching for a person named Brown. I have to go on giving more indexing to retrieve the particular Brown I am looking for. So that takes a large amount of time. I always see that people lose patience. You need better computing or better search techniques on a large scale. There's going to be a problem in years to come because there is lots of information and lots of redundancy on the Internet. So the question is how do you pick out those that you don't need? When you are searching, you collect a lot of garbage data. How do we, on the first search itself, read through a lot of information so that does not happen? When I talk about information it can be audio information, image, radio, all sorts of things. If I am talking about images then I need a much faster way of accessing perhaps not the image itself but maybe a complement of an image. That calls for doing a certain amount of dynamic scene analysis on an image, which requires more sufficient, or high-performance search organisms. That takes a little more time. So that's one of the impediments I see -- how can you make efficient search algorithms on the Internet? I hope to design high-performance computing and communications search organisms using intelligent agents.

UBIQUITY: Talk about some of the research that you are doing now.

IYENGAR: The Robotic Research Lab (RRL) at LSU is comprehensively addressing a number of fundamental problems. Members are performing advanced research in the areas of intelligent sensing and vision; spatial planning and navigation; asynchronous production systems; integration of knowledge sources in distributed systems; and architectures for the control of mobile robots. The laboratory has conducted important work on the problems of intelligent robot navigation, intelligent sensing, world-modeling using acoustic sensors, and real time expert systems for the control of mobile robots, to name a few at this stage. RRL investigators are presently exploring several new promising avenues of research. We have a group of researchers at Concurrent Computing Laboratory for Materials Simulations (CCLMS) doing world-class work in the area of material simulation. This group (Profs. Vashishta, Kalia and Nakano and a group of graduate students) were the first people to do billion atom atomistic simulation. This strong research group has annual funding close to 2 million dollars per year from various agencies such as DARPA, NSF, etc. So we have established a very strong research program on material simulations at LSU. They have sophisticated algorithms, but the interesting thing is that they use very large scale distributing computing for meta-modeling of nano systems. This group organizes a conference every year called the Mardi Gras conference. In the past they have focused on "teraflop computing for grand challenge applications" and recently they have focused on information technology structure for large-scale material simulation. Here this paradigm shift makes a lot of sense because these are not just computer scientists but their background is from physics, chemistry and material science, and physical chemists. Many of them have degrees in other areas and some are computer scientists too. This is what I call a very good intensive collaboration and multi-disciplinary effort in the context of paradigm shift.

UBIQUITY: How has your program at LSU involved undergraduate students particularly, but all students in the new paradigm?

IYENGAR: It has been excellent in terms of hiring students because the students are exposed to state-of-the-art computing. We also have a good infrastructure and a multimedia laboratory for students to use called the Web Technology Center. They don't have to be majors in computer science. Anybody who is interested in computing can be connected to this process. There are many software tools and boards for technical and non-technical students being developed. And then there is another area called multi-scopic computing.

UBIQUITY: Describe multi-scopic computing.

IYENGAR: In my opinion, in multi-scopic computing you essentially synthesize fragmented or computational scientific coding. Let's say you have five different small programs or algorithms. In multi-scopic computing we maybe take each of these small segments of code and synthesize it in the context of software engineering tools. Say I want to take all of these small components of coding and synthesize it very efficiently for anybody to use. That's another interesting aspect of it. You can view this type of computation from different points of view.

UBIQUITY: How is the group involved with teaching?

IYENGAR: That's another thing unique about this group. It's been able to teach global courses. There are students who are working on projects here with a university in Europe. They can take courses, they can talk to the instructor, and work on projects. In other words, we think that the location is not important. I can be in Europe and still take some courses at LSU if there is the infrastructure to communicate from this place to that particular university or agency.

UBIQUITY: You say it's unique to LSU?

IYENGAR: At LSU we have a very strong group of world famous computational material scientists doing state-of-the-art work jointly with the computer science department. This is a remarkable achievement in collaboration and interdisciplinary research in terms of this new computational paradigm shift that I am talking about. There is one thing I want to comment on here. In the beginning in the 1990s when these researchers came from National Argonne Lab, there was a lot of reluctance on our computer science faculty to collaborate with them. This was the reaction in most computer science departments in early 90s but it has changed drastically in most computer science departments across the country. They were not accepted readily because the computer scientists thought, well, this is not computer science research. But even though there was some opposition in the department I saw we were in this paradigm shift of computing. We were interested in interdisciplinary solutions to problems. We were interested in global solutions to real applications. So I looked at it and tried to convince my faculty that it was important for us to move in this direction. It is to the advantage of our own department to have this collaborative faculty. But right now if you study many computer science departments they've changed their name to College of Computing Science because they have realized all of these shifts. Every university across the country is changing in this direction. That's why I'm very hopeful that in the next 10 years, even though there are some problems with the digital divide, we can certainly be optimistic and see the results accruing to each and every person across the world.