Research Project Highlights

A Library of Simulation Optimization Test Problems

Stochastic simulation is one of the most widely used Operations Research techniques, being applied in almost every industry, with particularly prominent use in manufacturing, health care and financial engineering. This popularity arises because of its ability to capture essentially arbitrary model complexity. Stochastic simulation is almost always used to assist with some decision, and so we are naturally led to the problem of simulation optimization, i.e., optimizing some function that can only be evaluated through a stochastic simulation. A large number of simulation-optimization algorithms exist, but little effort has been made to compare them. Professor Shane Henderson is collaborating with Professor Raghu Pasupathy at Virginia Tech, along with a coalition of Ph.D. students and undergraduate students to develop a test-bed of simulation optimization problems. The test-bed, now available at www.simopt.org, includes approximately 50 test problems. Undergraduates at Cornell have helped create this website, create and write up the test problems, and implement the required simulation code for the problems.

Engineering an Interactive Laboratory Tour

We now have a number of distance learning courses and engage with practicing engineers at numerous off-campus sites. Software packages such as Adobe Connect and Skype make interactive sessions with off-campus groups quite easy but we are not yet fully utilizing their potential. A student project is presently investigating the challenge of conducting interactive, live, laboratory tours using web-based technologies. Allowing off-campus participants to ask questions will be an important ingredient. This would lead to a follow-on challenge of conducting live factory tours. Part of the project is “requirements discovery,” that is, through research, interviews, brainstorming, and experiments discover what makes for a successful live, but virtual, tour. Another part of the project is evaluating different commercial and homemade solutions to the challenge. A third part will be to define training and protocols for tour managers to ensure a successful tour. This project is directed by Professor Peter Jackson and run in cooperation with the Finger Lakes Chapter of INCOSE, the professional society for systems engineering, which has also expressed a strong interest in developing interactive factory tours.

Booking Limit and Pricing Controls in Airline Ticket Sales

Booking limits form one of the primary tools for airlines when they control the sales of airline tickets. In particular, booking limits put restrictions on how many tickets can be sold at each fare level, and by doing so, they allow airlines to keep a balance between not letting too many seats sold at low fare levels and not reserving too many seats at high fare levels. On the other hand, demand for airline tickets is clearly price sensitive. Revenue management models that try to capture the price-sensitive nature of demand suggest a price that should be charged as a function of time remaining in the selling horizon and as a function of the remaining number of seats on the flights. This situation potentially creates a mismatch between revenue management models that suggest a price to charge and ticket sale control systems that need booking limits for different fare levels. Several undergraduate students, under the guidance of Professor Huseyin Topaloglu, studied the problem of taking the price suggestions from a revenue management model and turning them into booking limits that can be implemented by the ticket sale control systems. They explored different approaches to make the conversion from price to booking limits and ran simulation experiments to see how frequently the booking limits should be updated to avoid significant revenue losses when compared with a pricing-based revenue management model.

Airline Crew Scheduling

For major airline corporations, costs associated with flights crews are second only to acquisition costs of jet fuel. Once flight schedules have been fixed, there are typically an astronomical number of ways in which the flight crews for given airlines could be allocated to cover the flight commitments. Choosing the best (i.e., least costly) among the possible crew assignments is a difficult large-scale discrete optimization problem. Sponsored by the College’s Engineering Initiatives, the undergraduate research program, three ORE majors worked with Professor. Leslie Trotter to study computational techniques for solving large crew scheduling models. This grew into a research project with over a dozen undergraduate participants studying computational software for linear and discrete optimization.