OR Snow Jobs

In my last post, I was grousing about being snowed in (Carnegie Mellon has been canceled three days and counting) and the need for more operations research in these sorts of situations.  I am pleased to see that my own university is taking up the challenge.  CMU President Jared Cohon has offered the services of the university to help the city create a state-of-the-art planning and operations system.  From an article in the Pittsburgh Post Gazette (and thanks to @bryanroutledge for pointing this out):

[Pittsburgh City Council Member] Mr. Peduto said CMU has offered to marshal faculty, staff and and students to create a “state of the art” snow removal tracking system that identifies priority areas for snow removal, maximizes the use of city equipment and follows through after emergencies with studies to guide budget decisions. It could also include geo-tracking of snow plows and treated streets, which is used in Maryland and elsewhere, he said.

The offer includes help from “the department of engineering and of computer science, the Heinz School” and other CMU departments, said Mr. Peduto, whose district includes the Oakland campus. “They’ve offered the full services of the university to create a better system.”

Snow removal has certainly been looked at by those in operations research. For instance, James Campbell of University of Missouri-St. Louis has been working on these problems for more than fifteen years. Even earlier, in 1976, Gene Woolsey wrote on snow removal in a delightful Interfaces article entitled “Three digressions on the routing of trucks” (the article is also available in the wonderful book The Woolsey Papers). In the article, Gene writes about giving a talk to city managers in Colorado:

After my speech, a city manager approached me with the following problem he had been facing for some time. It seems that the city council in his city had some years ago cleverly set the tim for council meetings at 6:00AM on alternate Tuesdays. The reason for this time should be immediately obvious, as it assures that only the citizen who is really upset will rise up to vent his spleen at that hour.

Unfortunately, when it comes to snow removal, lots of people rose early to complain. Gene’s solution?

Find out how the council members get to the council meeting from their homes, and clear those streets first. If you are not already doing this, and you do it in the future, I suspect that the complaints will go down.

That is really an OR snow job! If I had to guess, Pittsburgh discovered this plan years ago. This week’s storm suggests the need for a somewhat more all-encompassing approach.

6 thoughts on “OR Snow Jobs”

  1. one issue that would complicate pittsburgh route planning is all of our hills. i know people in greenfield who said the plow got stuck trying to get up their street. but it sounds like a fun and interesting problem – perfect for OR.

    i hope ravenstahl takes cmu up on the offer.

  2. Years ago I visited Long Island (which, for any culturally-deprived readers, is in New York). This was not long after LI got some sort of recording setting snow fall. (Proximity to the ocean normally moderates the climate there a bit.) They’d run into a problem I had no anticipated, despite 30+ years living in Michigan (no stranger to heavy snowfalls): they’d run out of places to put the snow. Snow throwers (and people shovelling manually) could not achieve the necessary lift to get fresh snow atop the existing piles, and plows were running into analogous problems.

    That leads me to think that optimal routing, scheduling and equipment allocation may be amenable to modelling under conditions of “ordinary” snowfall, but modelling the trapped-at-home-reading-journals variety of snowfall may be considerably trickier.

  3. Paul’s experience on “Lon Goy Lind” makes me think of that classic Intro Networks problem where the job is to move a minimum volume of earth along a road under construction (from the high spots to the low spots) to level the road bed.

    I suppose in Pgh the solution would be a lot more readily implementable than on LI (which is pretty damn flat).

  4. Actually, it should be easier to implement the solution in LI, which is 23 miles or so maximum width (north to south) — so you’re never more than a dozen miles from a convenient place to dump snow (Long Island Sound to the north, Great South Bay or Atlantic Ocean to the south).

  5. There’s at least two distinct parts in which OR can help for snow plowing. One is strategic, and is the dimensioning of the fleet. The other is tactical, and is the planning/routing of the plowing. Now, having a fleet of vehicle which, necessarily, depreciate, induces a capital cost. If we simplify and assume dimensioning is a one variable problem, with the variable being that capital cost per year, the optimal fleet is likely the one were one additional dollar invested in the fleet decreases the expectation of the loss in productivity of the city due to snow by one dollar.

    I don’t know how much productivity was lost in Pittsburgh (or in the north-east in general) due to the slow snow removal, but one thing is sure: given that is was the 4th largest snowstorm in 130 years, if all the streets had been perfectly cleaned the day after, that would have clearly meant that the fleet was oversized — hence inducing capital cost to the city that are larger than the expected productivity they save. So I see no contradiction between “using OR” and “having to wait 4 days to have one’s street cleaned”.

    Now, this reasoning is purely strategic, and says nothing about the operations, were there is likely room for improvement.

Leave a Reply to Paul Rubin Cancel reply

Your email address will not be published. Required fields are marked *