There are significant hurdles to providing real problems and data for O.R. modeling and algorithm development to the research community:

• Industry data is a valuable asset. Companies competing to make a profit do not want to "give away" anything that might aid their competitors.
  
  
• Defining a problem to share can take considerable time and effort.
  
  
• Terminology and concepts used by industry may be unfamiliar or understood differently by the research community.
  
  
• The time and effort required clearing the bureaucratic hurdles of management and legal approval to share a problem or data can cause the best-intentioned efforts to fail.
  
  
• Lack of clarity in export controls, such as the International Arms and Traffic Regulations (ITAR), leads corporations to err on the side of non-disclosure rather than risk fines and imprisonment.
  
  
• Collecting and formatting data for others to use can take considerable time and effort.
  
  
• Urgent industrial problems often need to be addressed over a time-span that is much shorter than deep research requires.
  
  
• The goals of the application provider (e.g., to obtain a workable solution) and the goals of the research provider (e.g., to create original, publishable scientific results) are often at odds.
  
  
• Whatever channel is used for distributing real problems and data must be maintained and user questions must be answered.

Despite the hurdles, successful partnerships between application specialists and academic researchers exist (e.g., see past Edelman winners). Typically, the partnership is between one corporation and one university. Relationship and intellectual property issues are resolved in advance. The phenomenal impact of the partnerships showcased by the Edelman finalists prompts the question of how to replicate these types of successes. Unfortunately, the typical model of engagement does not scale well.

Can we reap the benefits on a larger scale by making real-world application problems and data available on-demand to the universe of O.R. researchers, teachers and students under pre-defined terms of engagement?

At the recent INFORMS annual meeting in San Francisco, a panel convened to discuss the challenges of establishing an open repository of real problems for modeling and algorithm development [2]. The panelists included representatives from AMPL, Boeing, General Motors, IBM, Lakeside Associates, Lucent Technologies, NSF, UPS and the University of Arizona.

Drawing from their own experiences of working with external research organizations, the panelists expressed a wealth of opinions, frustrations, challenges and questions:

• What would be the expected gain for contributors? Is it worth the investment in time and energy for companies to participate?
  
  
• What would a problem or data contribution to the archive look like?
  
  
• How would a library of real problems and data work?
  
  
• Is there any bottom-line, short-term benefit for the contributor, or is this mostly for the "good of the profession?"
  
  
• Are there opportunities for companies in the same industry to cooperate in establishing archetypical "real" problems and data for their mutual benefit?
  
  
• Are we at a "tipping point" with current technology that some of the past roadblocks can be overcome (e.g., XML for data formats)?
  
  
• Is there government support for establishing and maintaining this type of infrastructure initiative?

Three action items resulted:

• Write this article for OR/MS Today to generate awareness and solicit feedback.
  
  
• Establish a Web site and mailing list to foster the discussion of the problem exchange. See www.coin-or.org/RPX.
  
  
• Plan follow-up events for the practice community at the next conference opportunity.

There will be multiple opportunities at the 2006 INFORMS Practice Meeting (April 30-May 2 in Miami) to discuss and help define the form, content and operation of the Real Problem Exchange. During lunch and the "birds-of-a-feather" sessions on Monday, and again during lunch on Tuesday, there will be facilitated breakout sessions with two related but different goals.

One set of breakout discussions will focus on defining the form and characteristics of problems to be shared in the Exchange. The facilitators will be drawn from the practitioner community and will come with example problem statements, models and data sets. To stimulate discussions, a number of pre-selected questions will be distributed to meeting participants prior to the event and posted on the RPX Web site. For example:

• Should the problems be company-specific, industry-specific or as generic as possible (recognizing that companies would like their specific problem solved but that academics gain tenure by publishing papers that are as broadly applicable as possible)?
  
  
• Can the problems be stated in such a way that they are clear but solution-free (because we hope to foster multiple solution approaches to each problem in the set)? Or should the problems be stated as a challenge with the possibility of a best answer?
  
  
• Can the data sets be supplied in a compact form yet still be used to generate a large number of problem instances?
  
  
• What would motivate the research community to provide solutions to the problems?

Another set of breakout sessions will deal with the policies, procedures and systems required to operate and maintain the Exchange. Again, there will be facilitators and pre-selected questions. For example:

• Should anyone be able to deposit any problem in the Exchange? Should there be a strict format? Should there be some review before insertion?
  
  
• How much support from the author of the problem will be required by interested researchers? What kinds of support mechanisms should be available (e.g., telephone access, e-mail access, question-and-answer areas on the Web site)?
  
  
• Should a dictionary of terms be provided to ensure terminology is well understood between the research community and industry?
  
  
• Should solutions be posted on the Exchange or published in the literature or both? (It would be useful to always have a pointer to the best-known solution to each problem.)
  
  
• How can the Exchange be supported? Should the posers pay each time they insert a new problem, or only when a superior solution is supplied, or should there be a paid subscription mechanism? Should there be corporate and institutional sponsors?

The facilitators will consolidate the results of these discussions and produce a draft set of guidelines describing: 1) the form and properties of the problem statements, models and data sets that will be collected and distributed in the Exchange, and 2) the policies, processes and systems that will be required for efficient operation and maintenance. These results will be distributed and published to stimulate further discussion.

In addition to producing the guidelines, the practice meeting discussions will help identify interested industrial and academic people who are willing to invest their time in turning the Real Problem Exchange idea into reality.

The Exchange is just one actionable approach to the larger question of how to strengthen collaboration between O.R. practice and research communities. We welcome your comments on the Real Problem Exchange idea and related topics. Join the discussion by subscribing to discussion mailing list at www.coin-or.org/RPX.

Acknowledgements We are grateful to Harlan Crowder, Abhi Deshmukh, Debra Elkins, Bob Fourer, Michael Grant, Dave Heltne, Ranganath Nuggehalli and Suvrajeet Sen for sharing their views and provoking this article.

1. Suvrajeet Sen, "Bridging the Gap," 2006, OR/MS Today, Vol. 33, No. 1, pgs. 30-33.
   
   
2. Abhi Deshmukh, Debra Elkins, Bob Fourer, Michael Grant, Dave Heltne, Robin Lougee-Heimer, Ben Lowe, Ranganath Nuggehalli and Suvrajeet Sen, "Creating a Testbed of Industy Problems for O.R. Model and Algorithm Development," INFORMS New Orleans/San Francisco, November 2005.

Table of Contents

OR/MS Today Home Page


OR/MS Today copyright © 2006 by the Institute for Operations Research and the Management Sciences. All rights reserved.


Lionheart Publishing, Inc.
506 Roswell Rd., Suite 220, Marietta, GA 30060 USA
Phone: 770-431-0867 | Fax: 770-432-6969
E-mail: lpi@lionhrtpub.com
URL: http://www.lionhrtpub.com


Web Site © Copyright 2006 by Lionheart Publishing, Inc. All rights reserved.