Profiles in OR/MS: Ladislav Lettovsky
|
Senior Consultant,
B.S. Cybernetics, University of | PhD Industrial Engineering, Georgia Tech
Contact Information: |
Questions & Answers
Q. Tell me about your academic and professional background and how you became interested in the field of OR/MS?
A. I got my undergraduate degree in Cybernetics from the University of Zilina in Czechoslovakia and took some O.R. courses as part of the core curriculum. One of the courses I particularly enjoyed was Graph Theory and Algorithms. During this time I focused on problems related to the Rail Industry. After graduation, I went directly into industry and worked for a company in Czechoslovakia that developed railway-scheduling software.
During this time, the computer industry was evolving at such a fast pace. My job was technically challenging and applied much of the OR theory that I had learned as an undergraduate, but I felt I was missing out. As the speed, power, and memory of computers grew, I felt I was losing the edge. So, I decided to go back to school. I got a scholarship to attend the University of London where I received my Masters in O.R.
Nearing the completion of my Masters degree, I was tempted to enter academia because I wished to gain a more theoretical background. I had become interested in integer programming and decided to take a research position at the University of Loughborough where I investigated how to apply genetic algorithms within branch and bound. After 3 months, the University was still trying to get a work permit for a full time research position and I was beginning to get nervous. Fortunately, I soon received a letter from Professor George Nemhauser to come to Georgia Tech to study and do research with him. I had previously written to George expressing my interest in his integer programming book and in my doing research with him. So, I left London for Atlanta and enrolled in the PhD program at Georgia Tech. At Georgia Tech, I worked with Professor Nemhauser and Professor Ellis Johnson on real-time algorithms for Airline Crew Schedule Recovery.
In the summer of 1994, I applied for a summer internship with US Airways, but was instead offered a full-time job. I accepted the position with US Airways where I worked on the development of systems for Crew Schedule Recovery and Dispatcher Workload Assignment problems. After nine months at US Airways, I heard that Sabre (at that time, Sabre Decision Technologies) was forming a Research Group. I was interested because it sounded like the right combination of research and application. I took the job with the Sabre Research Group where I continued my work on real-time Crew Schedule Recovery. I also began investigating algorithms for real-time Aircraft Recovery and, all the while, was still working on finishing my thesis for my PhD.
Q. How difficult was it to finish your PhD thesis while working full-time?
A. I do not recommend it because it was extremely hard to stay focused. I would say that if someone is tempted to leave school before finishing their PhD, they should only consider doing so if what they are going to do in industry is exactly the same thing as what they are researching for their thesis topic. I was lucky because working in industry on Crew Recovery (my thesis topic) made the thesis even better (I could no longer made any artificial assumptions). I had to make sure that I came up with was useful and applicable. You can get a lot of that by working with a real client.
Q. What projects are you currently working on at Sabre?
A. We are now working on a framework for an Integrated Airline Recovery System for real-time day of operations aircraft, crew and passenger recovery. This will more or less put together the work we have already done on each resource independently. But now the challenge is to come up with a unified solution that is acceptable to everyone. If you find a solution that at least one of the users can not live with, no matter how optimal it is, it won’t be implemented. Historically airlines have divided their day of operations personnel (e.g. aircraft routers, crew schedulers, maintenance routers) into distinct silos often with conflicting goals. Thus, determining the objectives for an optimal unified solution is not a trivial task. In addition, there is little communication across the silos. So introducing integrated recovery will require a change in the airline’s business practices before it will be fully incorporated within the airlines.
Another project I have recently worked on is Flight Schedule Generation, where we have come up with a way to build a new flight schedule either from scratch or using the current schedule. From working in this area, I have found that the airlines are using optimization very heavily in the planning stage. As a result, the schedules produced are very tight. There is very little slack left for disruptions because the optimization is good at cutting out the slack! So we end up with a flight schedule that according to the objective function is optimal, or near optimal, meaning that there should be big savings, however such savings are not realized in practice. In fact, applying optimization too heavily may result in airlines losing money. That is why it is so important to have good recovery algorithms, algorithms that repair disrupted schedules in real time.
I think that the next step is move towards combining the scheduling and day of operations work and do robust scheduling. In other words, how to create a schedule that will implement well on day of operations. I believe this is the next frontier of airline research in the area of flight scheduling. Not many people are currently working in this area.
Q. What do you see as the primary role of the Research Group at Sabre?
A. To maintain the leading edge of Sabre by investigating new technologies and algorithms for application in our current and future products. We provide proof-of-concept for new products and work to enhance existing products. The Research Group at Sabre is very focused on application. With very rare exceptions do we ever do something for which we don’t see immediate applicability.
Q. What do you like most about working in the Sabre Research Group?
A. The interdisciplinary composition. In spite of the relative size (there are currently 15 people), we have people from many different fields who are real experts in their respective fields. And, because of the flat organizational structure (of the research group), when there is a problem it is very easy to talk to someone else, or pull a couple of people together to work on a problem. It is very rare that you will find a problem where only one person’s expertise will be sufficient, because most of the problems that we are dealing with are of an interdisciplinary nature. For example, most of the problems we tackle require a forecasting component, may require implementation within a parallel architecture, and usually involve a combination of traditional optimization and artificial intelligence techniques...and we have experts in all these areas. Also, we have people that are also very good at data visualization so that we can represent the results of our work in a meaningful form. So that is what I like most about the Sabre Research Group – that we have real experts from many different fields of research and it is very easy to bring a team together to work on any of our problems.
Q. What advice do you have for those just starting out in the field of OR/MS?
A. As far as technical skills that are important, a combination of computer science (e.g. data structures, algorithms) and operation research skills are important because nowadays everything that you do you need to take at least to the proof of concept stage. Even though there are now a lot of modern, high-level, languages, you still end up doing a lot of coding. And, even with increased speed and memory you will quickly exhaust the capabilities of the computer.
Q. You have studied in both Europe and in the US. Have you noticed a difference in the way OR/MS is taught?
A. I studied at the London School of Economics in Europe which had a “soft O.R.” focus whereas the schools in the United States are more “hard O.R.” oriented.
In Europe, my studies were more focused on solving the problem. We are taught that it is okay to use the simplest technique possible if it solves the problem. In fact, the preferred technique is the simplest technique that will solve the problem. Many times projects will end up in disaster when people feel that they need to throw all their "algorithmic machinery" at it.
The schools in the US are more focused on the mathematics and theory behind the O.R., which I also find very interesting.
Q. How do you continue to expand your knowledge of new technologies and new techniques in OR/MS?
A. I go to conferences to find out what is new. Usually as an outcome of going to a conference, I end up reading one or two papers to learn some more. From conferences I get a little insight, but I usually do not learn that much up front. I usually have to get the paper and read the article to gain a more in-depth knowledge. What I do get from conferences is a feeling for where the industry is heading, what the current problems are, and what is now considered solvable from an industry perspective.
I also subscribe to several email lists to get information on upcoming conferences, new papers, and new books.
Q. You are a member of AGIFORS and also the Chairman of the Operations Control Study Group. What is AGIFORS? How has it benefited your work?
A. AGIFORS is the Airline Group of IFORS (International Federation of Operations Research). The organization supports various study groups in specific areas (e.g. yield management, cargo, operations, crew scheduling, and flight scheduling) as well as an annual symposium. At the annual symposium all the study groups come together along with airline managers, vendors and research centers. The best papers and talks from each of the study groups are presented and vendors set up booths to market their products. This benefits all participants as it fosters an exchange of ideas.
Q. What do you predict the future has in store for the field of OR/MS?
A. I think there will be more work – more and more opportunities in the future. Take yield management as an example. Initially just the airlines began using yield management concepts to increase their revenue, now it is used just about anywhere – hotels, trucking companies, car rental companies, and so on. Also O.R. is being used more and more in Supply Chain management.
I also see O.R. being used more in a real-time environment. I think that we will see traditional optimization techniques being used more in combination with heuristics to find close to optimal solutions to NP hard problems in real-time. And we will continue to push computers to the edge of their computing power.