RAS Problem Solving Competition

Winners are now announced here
Announcing:
The 2018 Railroad Problem Solving Competition
Advanced analytical approaches applied to real-world rail problems
First Prize: $2000 --- Second Prize $1000 --- Third Prize $750
Predicting Near Term Train Schedule Performance and Delay

>> Detailed Instructions and Data Now Available

Description:

Railroads have an intense level of congestion owing to high volumes, large trains, and highly constrained track infrastructure.  Often, a relatively minor event such as a late crew or defective car can cause large ripple effects throughout a rail network, causing congestion delays to a large number of trains. Often, the view of the delay is based on a single train, though often a train’s delay is the result of problems created by other trains and events in the rail network.

In the Netherlands passenger train transportation is a very important mode of transport, with over a million passengers travelling by train every day. Netherlands Railways is the biggest passenger train operator and operates almost 6,000 trains daily. ProRail is the infrastructure manager, responsible for track maintenance and coordinates train operations for all operators.

For smooth punctual operations reliable real time data is of the utmost importance. Over the recent years real time data regarding delay in the entire network has been collected and made available to the operators. This data is used for passenger information systems and in dispatching centers.

Competition Objective:

Netherlands Railways and ProRail would like to have accurate forecasts for the rail network on ultra-short term, say 20 minutes ahead. Currently, the most commonly used estimate for future delay is that the current delay remains unchanged. Obviously, this is often not the case. A good estimate for what the delays will be over the next period of time will benefit both passenger information and dispatching of the railway operation.

The goal of this competition is to create better forecasts of train performance and delays. Entrants will apply any appropriate advanced analytical techniques to better predict near-term train performance.

No rail experience necessary!  A full, detailed description of the problem is forthcoming, and questions from participants welcome!

Data:

Comprehensive data on train deviation from planned schedule will be provided from ProRail for contestants to use advanced analytical techniques to forecast future delays, including planned and realized train timetables, planned and realized crew schedules and planned and realized rolling stock schedules.

A sample of the time table plan and realization is available at the competition website for those interested in previewing a subset of sample data. Other more complete data sets and descriptions are forthcoming.

Competition Criteria:

The criteria that will be used to evaluate a solution include:

  • The accuracy of the solution – How accurate is the predicted train performance?
  • The soundness and sophistication of the solution approach –
    How advanced is the methodology, and how robust is its application?
  • The quality of the paper describing the approach – How clear is the explanation?
  • The quality of the presentation, to be given by three finalists at the Rail Application Section Meeting at the INFORMS conference in Phoenix, Sunday, November 4-Tuesday November 7. (The attendance and presentation of at least one team member required to win.)

Competition Awards:

  • First Prize: $2000
  • Second Prize $1000
  • Third Prize $750

Apart from the cash prizes, the first prize winner’s contribution will also be considered for publication in the journal Networks. The winning will receive an expedited refereeing and publication process.

Competition Calendar:

  • Registration: Deadline is April 30, 2018
    Interested parties should register by sending an email to:
    railwayapplicationssection@gmail.com
    Your email should include all team member name, contact information, and affiliation.
  • Full Problem Release:  February 28, 2018
    Additional data, more specifics of the problem
  • Question and Answer period:  February 28 – July 1
    Participants may ask questions; all questions and answers available to all participants.
  • Quiet Period: July 1 - Aug 1.
    Participants may continue to work on solutions; no additional information provided.
  • August 1 – Solution due
    Solution includes report on methodology, and solution data set
    (format of solution data set provided Feb 28)
  • September 1 – Finalists announced
    Finalists are expected to give a presentation at INFORMS conference,
    Phoenix, November 4, 2018
  • November 4, 2018, Phoenix - Finalist Presentation.
    Each finalist gives a 15-20 minute presentation on their approach;
    Judging panel asks questions.
    All finalists must attend.
  • November 4, 2018, Phoenix Winner Announced

Good luck in the competition!

Professor Michael F. Gorman, Ph.D.
University of Dayton