Wednesday, January 5, 2011

Operations Research from Train Scheduling to Formula One Racing!

I was delighted to read about the research of Professor Warren Powell (thanks, Warren, for the great holiday letter, by the way) and his team at Princeton University for Norfolk Southern Railroad in Wired Magazine. Their Plasma (Princeton Locomotive and Shop Management System) uses approximate dynamic programming to enhance decision-making associated with Norfolk Southern's operations, and predicts the impact of changes in fleet size, maintenance policies, transit time, and other key factors.

The article on Transportation begins with "Model trains are easy to keep track of. But building a model to run real trains is a complex undertaking." Nevertheless, I could not resist featuring some of my husband's model trains, since he is an avid model train collector (as was his Dad), and some displays from the Amherst Railway Society annual show.

Now, if the above hasn't whetted your appetite for algorithms and computer-based mathematical models in transportation,The New York Times is reporting that Ferrari hired Neil Martin as its Head of Operations Research. Neil Martin had formerly worked for Red Bull and was head of strategic operations for the Red Bull Formula One racing team. He has a degree in mathematics and computer science and another one in operational research (what the British call operations research).

In a fascinating writeup a while back in The New York Times, Martin was quoted as saying: "Before I got seduced by Formula One, I was headed right into the city doing derivatives and share trading," he said. "And in a sense, this is completely analogous to it."

"All we were doing was like in a financial portfolio, spreading the risk," he said.

The article continues: But such decisions are based on data analyzed through millions of scenarios on computers that work not only throughout the week before the race, but during the race weekend and the race itself, as the team's results and those of the other teams are added.

Such strategy started to become crucial in the mid-1990s, when refueling and tire changes were introduced, meaning that a car no longer ran an entire race on a single tank of fuel and one set of tires. The importance increased in 2003, after cars were required to start the race with the same amount of fuel that they had used in qualifying. That meant deciding how much fuel to use based on an expected grid position and on the ideal length of the first stint.

The main factors fed into the computers are number of race laps, amount of fuel a car should carry, the car's speed and the duration of its tires, and what the competition is expected to do.

Isn't transportation fun!