M. Eng. Students at Work on a Challenging Set of Projects
As spring slowly emerges in Ithaca, M.Eng. student teams are entering a crucial phase in the development of their project results. Projects that deal with real problems encountered by real organizations have been a cornerstone of ORIE's M.Eng. program for more than 40 years.
Although most Financial Engineering projects take place entirely in the fall semester at Cornell Financial Engineering Manhattan, projects in other Concentrations -- Applied Operations Research, Information Technology, Data Analytics - as well as the Manufacturing and Industrial Engineering Minor - begin in October in Ithaca and continue through the winter break and spring semester. Read on for detailed information about these projects, together with one Financial Engineering project that is based in Ithaca and is taking place this spring. (There is also project work within the Strategic Operations Concentration courses and in the Systems Engineering Minor that is not discussed in this article).
The project slate students are engaged in this spring is typical of recent years, incorporating projects in a particularly broad array of subject areas: telecommunications, agriculture, transportation, internet marketing, health care, disaster planning, emergency medical services, manufacturing, and financial engineering. They are listed in the following table, with links to more details about each project from the short project title, together with links to the sponsoring organizations and faculty advisors.
Last year's project for Canadian National (CN) Rail led to a journal publication. This year's CN Rail team is working on a classic Operations Research problem - developing schedules for work crews that match proposed train schedules. This problem is routinely solved by airlines, but the rail version has its own character, according to team advisor Professor Peter Jackson. Like the airline problem, train crew scheduling is computationally intensive, entailing the solution of large integer programming models, an area in which co-advisor Professor Leslie Trotter is an expert. It also has to take into account random events, such as unplanned absences and arrival delays.
|CN team members Christian Klar, Mikhail Chrestkha, Abhas Sinha and Betsy Wang explain their poster to prospective M.Eng. student Eric Webb (center) from Case Western.|
The CN Rail team presented a poster about their project at the annual Cornell Engineering Research Conference on April 3. Poster presenters have the challenge of distilling research projects into a single large page and being prepared to use it to explain the project to attendees of the poster session. With the objective of providing a tool for train dispatchers to use in minimizing the cost of crew assignments, the team is employing integer programming optimization techniques, discrete event simulation and Monte Carlo simulation.
The yield from a farm field depends in part on the depth of the top soil. Nature tends to redistribute the top soil over time, and the farmer is faced with the challenge of rearranging the top soil with the objective of maximizing yield. However rearrangement, which is accomplished with a tractor-drawn scraper, can be time consuming and expensive. To farmer and Cornell agricultural engineering Ph.D. student Clay Mitchell, this suggested that Operations Research might be useful. He first conceptualized the problem for a term paper in an ORIE class on applied systems engineering taught by Professor Huseyin Topaloglu. After graduating with a degree in biological engineering from Harvard, Mitchell had "bought 200 acres that are a long way back on a dusty stream following road where old trees stand as mileposts to the bygones," according to his web site. On the Mitchell Family Farm in northeast Iowa he has been combining traditional farming sensibilities with modern automation on those 200 acres, an experience that brought him to Cornell for Ph.D. studies.
|Shaan Qamar hands the marker to Anshuman Bhairavbhat to augment the model formulation, as Professor Huseyin Topaloglu, Talha Omer and Kevin Ham look on.|
A somewhat unusual M.Eng. project positioned on this farm has grown out of Mitchell's term paper. The project entails determining how yield depends on depth and then determining where and how topsoil should be redestributed to balance the tradeoff between cost and benefit. Working from maps of relative yield and data about soil depths, the team is attempting to develop software that optimizes the redistribution of topsoil at each point in a grid superimposed on the field, taking into account the workload involved in moving topsoil from one place to another. The problem has elements of the classic traveling salesman problem (in which a sales person tries to visit her customers in an optimal order) but entails working out how much to remove or deposit at each grid enroute. Moreover, yield is not simply proportional to top soil depth, but has diminishing returns as more soil is added. "At each point in the field there is a 'sweet point' where marginal benefit = marginal cost for each movement of the tractor," according to team member Shaan Qamar.
"I personally feel lucky to have been part of this particular M. Eng. project, " said Qamar. "It is rewarding to explore problems to which there is no 'right' or 'wrong' approach. The client, Clay Mitchell, has assisted us in understanding that it is difficult in the farming industry to add value immediately [but] our initial results have shown that small increases in profitability can add up to something substantial in the long run." The project promises to plow new ground in the application of operations research techniques.
|Professor Peter Jackson and team member Nisha Bangera review approaches to implementing risk assessment software.|
This year's project for Cayuga Medical Center, the hospital that serves the Ithaca area, is assessing the potential impact of a pandemic influenza outbreak on the hospital's operations. The team is using computer simulation to understand how various policies governing the deployment of human and material resources over time will perform under the time-dependent load resulting from an influenza pandemic. The are building a software tool that hospital administrators can use in assessing risk. Public Health Professor Nathaniel Hupert of Cornell's Weill Medical College has provided data and insights to the team. He is co-director with ORIE's Professor Jack Muckstadt of the Institute for Disease and Disaster Recovery and was recently named head of the Preparedness Modeling Unit of the Centers for Disease Control.
"This project made me appreciate the various aspects of the health care system's supply chain, and the significance of emergency planning," said team member Nisha Bangera. "It became necessary for us to understand what different users would want from the tool, and how to make it as user-friendly as possible. We had not expected that it would be just as hard setting up the math and running the simulations," she added.
|Kathuri Rangan Prabhakaran shows the network currently in use in France by a company, and the network proposed by the team. Their proposal cuts the cost by more than half, to less than 3/4 of a million Euros per month.|
Global Capacity considers itself to be "the world's first telecommunications logistics company." The company keeps track of telecommunications service providers and pricing plans in more than 170 companies. This information is used to assemble services from various providers, who may base their pricing on different attributes, into a least-cost network for use by a large enterprise. Team members said that the project "made us realize the importance of the tools we learned during our courses and has given us confidence that we can do well in industry."
|The Global Capacity team holds regular weekly electronic meetings with the client. Professor David Williamson, Aayush Mahendru, and Tonghe Shen are intent on the client speaking to them through the teleconferencing device.|
The Global Capacity team was selected to give a presentation in the Innovation and Technology session of the April 3 Cornell Engineering Research Conference. In the presentation, team members showed successively more difficult versions of the specific problem they solved, for a network in France. The most complex version of the problem, which introduced a constraint that locations can only be served from within the 'province' where they are located, turned out to simplify the computations significantly.
ORIE M.Eng. students have been working with Analog Devices, Inc. (ADI) for several years, assisting the company in improving the quality of traffic to their web site. In past years project teams have influenced ADI's strategy in bidding for Google Ad-Words, on maximizing the flow of traffic to the ADI sites, and to increasing the effectiveness of the traffic flow in producing business for ADI. This year's team is determining which design and layout attributes of ADI's web pages impact converting site visitors into customers.
ADI Team members have commented that "it is definitely interesting to see the growing trend [towards online marketing] and to try to help ADI change their website to improve brand awareness and increase customer activity." They note that "working with a real company is extremely different from working onan assignment for a course." Unlike well structured homework assignments with specific answers and due dates, client "needs and demands change, meaning the project changes along with it," team members said.
|Jared Gearhart points out a property of the abstract representation of evacuation routes to team mates Rui Li, Ahmed Bajwa, and Liuquan Han, Professor Mark Lewis, and MITRE clients Eric Blair and Rick Rudman.|
MITRE Corporation, a not-for-profit organization chartered to work for the public interest, has sponsored a series of M.Eng. projects. The latest project, building on work from a 2007-8 team, is considering the problem of managing the evaculation of a large urban area in the event of a catastrophic disaster. They are studying a scenario in which Washington, D.C. and its surrounding counties in Virginia and Marlyand are evaculated to West Virginia. The team is using both optimization and simulation techniques to gain an understanding of the resulting traffic flow, with the aim of determining what resources such as water, food, blood plasma should be stocked, and where.
To the six emergency medical services serving 113 square miles southeast of Rochester, New York, assuring 24 x 7 coverage is a major challenge. The services operate land vehicles, primarily with volunteers. They fall under an organizational structure called The 3rd Battalion. Often service requests to the Battalion, totaling 15,000 per year, come in with little time between calls. An M. Eng. project team is analyzing data and developing scheduling algorithms to make the best use of limited staffing to reduce response time, which can be a matter of life or death.
|Team members Ying Xian, Johannes Essl and Jong Yub Chae meet with Professor Shane Henderson to discuss approaches to the ORNGE project.|
Although the Canadian province of Ontario lies just across Lake Ontario from Rochester, new York, emergency services there face different challenges, are organized differently, and have different equipment than Third Battalion. Ontario covers 1 million square kilometers, of which 90% is wilderness, so helicopters and fixed wing aircraft are used to carry about 18,000 patients per year. The provincial government contracts with an organization called Ornge to provide this service, which includes both transferring patients between medical facilities and responding to emergency calls. Last year, an M. Eng. team developed an approach to determining the best locations to base the service, using both optimization and simulation techniques. Building on this award winning work, the current team is looking how Ornge transfers patients between medical facilities on a non-emergency basis. Such transfers, which can be planned ahead, constitute the bulk of Ornge's business. At the end of each day, Ornge flight planners construct routes to cover the next day's requirements at minimum cost. The project team is automating this process using integer programming. Their work product should free up flight planner time and is expected to turn up route solutions that are difficult or impossible for flight planners to identify with manual techniques, according to Professor Henderson.
|Professors David Matteson and David Ruppert point out an aspect of the housing data to team members Yasin Khan, Wesley Tillu, Vikas Garg, Yi Tang, Shirley Lu and Tim Roberts.|
In earlier years, the connection between housing prices and financial engineering might have been viewed as tenuous, but that connection has become headline news. Although most financial engineering projects are carried out in the fall semester at Cornell Financial Engineering Manhattan, ORIE's office in New York's Wall Street district, some financial engineering students are not able to schedule their project work in the fall. As a result a project team is working this spring to refine statistical techniques for predicting the price of homes in each state as a function of variables such as foreclosures, availability of credit, population size, issueance of building permits, and other available information. A particular characteristic of housing prices is that they have "long memory," which creates an interesting challenge for modelling, according to the team.
The ability to accurately forecast housing prices, together with an understanding of options pricing and portfolio management, is essential to valuing assets such as mortgage backed securities and credit default swaps. Speaking for the team, Team member Tim Roberts notes that "this project has been a great tie-in with financial engineering course work, particularly Time Series Analysis, Statistics for Financial Engineering, and Stochastic calculus." The team has learned how to deal with "issues that routinely crop up when using real-world data from multiple sources. The project has been an immense learning experience for everyone," Roberts said.
Pall Trinity Micro, a subsidiary of Pall Corporation, manufactures medical filiters at a plant in Cortland, New York, less than 20 miles from Cornell. The Manufacturing and Industrial Engineering Minor team has been visiting the plant regularly in a project to develop a "lean manufacturing" plan to improve the process of making the filters. The team is employing a technique called value stream mapping, which is typically used to identify opportunities to reduce manufacturing lean times. Although Manufacturing and Industrial Engineering is classified as a Minor, meaning that students from more than one engineering discipline can participate, this year's team is entirely made up of ORIE students.