Poofy Goof Rant-O-Matic

"The Railway Applications Section (RAS) provides a forum to bring together practitioners, consultants, and academics interested in applying operations research and management science (OR and MS) techniques to the railroad industry."

they're running a contest: http://informs-ras.org/Problem.htm the problem is to find an optimal fueling system for a provided rail network. the trains can't run out of fuel between yards, and the railroad wants to minimize cost. find the cheapest solution!

it's tempting to spend some time to write a fueling solver for the railroads. there is some reward money. how hard could a heuristic search be? create the basic data structures in C, convert the spreadsheet data into CSV files and write the parsing routines, write train movement simulations, and then start iteratively fiddling with fueling locations and fuel amounts.

I can't imagine each iteration would take too long to run. 214 trains, 73 yards, 214 locomotives, how hard could it be?

I can think of a few heuristics off the top of my head:

run the simulation and see what the last yard each locomotive hits before it runs out of fuel, and put a truck there. start moving fuel trucks backwards in the train routes and find out where the minimum number of trucks is before trains start running out of fuel.

start placing trucks at the cheapest fueling locations and keep adding until trains stop running out of fuel.

find out the most active yards and put fueling trucks there. if trains still run out of fuel, apply another heuristic.

this doesn't even touch on combining or chaining the heuristics.

Steve got burned out on linear optimization and minmax problems from the financial industry. I still think they're nifty, and any opportunity to learn about rail infrastructure is pretty cool to me.