An article in Nautilus by Tom Vanderbilt [Link] is about the intersection between optimization and human behavior. It is entitled “Unhappy Truckers and Other Algorithmic Problems” with the subtitle “Algorithms: Transportation optimization starts with math, but ends in understanding human behavior.” That pretty much says it all. The article is mostly about how optimization is used to direct human behavior, such as routing delivery schedules for truck delivery services (e.g., UPS), generating airline crew schedules, etc. What is “optimal” from a mathematical point of view is not always pleasant from a human point of view. Humans get burned out by efficiency.
When a [UPS] driver stops the van, he “has nine seconds to select a package and get out of there.” His tone suggested he was talking about a member of a bomb disposal unit.
Vanderbilt goes into an explanation of the TSP at length (see my last post) as well as Bill Cook’s and Warren Powell’s work. While I thought the article had an unfairly negative tone on the role of optimization in managing large-scale systems (optimization = trucker torture?), it correctly characterized the richness of real-world optimization problems. Specifically, Vanderbilt discusses the real-world constraints imposed by union rules and the challenges this brings to the optimization side. I liked this quote: “Trucks are simple; drivers are complicated.” In the end, this was a nice article about the challenges inherent in implementing optimization results in the real world.
Link to the full article here.
Punk Rock Operations Research 
July 19th, 2013 at 8:44 am
Doesn’t this remind you of the adage that it is easier to schedule machines than people? And that to implement OR models, you need a mixture of politics, diplomacy and psychology?
A long time ago, before the PC offered easy-to-access routing problems, I attended preentation by a company truck scheduler for a company delivering to large grocery stores. They had used heuristics for the routes they used, but one driiver was never assigned to deliveries for one store. Why not? That driver had a mistress living close by, and on the rare occasions when he had been assigned to deliver to the store, he and the truck had gone AWOL for a couple of hours.
We have several toll bridges in the UK where there is a charge in one direction and not in the other. (This is, of course, an OR solution to the problem of collecting tolls efficiently.) It has been noted that many truck drivers will drive from A to B over the bridge in the “free” direction, but will not when going from B to A and needing to pay the toll. The extra fuel and time more than covers the toll charge. But drivers do it, on the basis of not wanting to have the congestion at the toll-booths, or the need to carry a means to pay, but above all, because they would avoid the tolls when driving their private cars.
July 19th, 2013 at 9:08 am
It really boils down to having a solver implementation which makes it easy to add extra constraints – based on business users’ feedback.
July 19th, 2013 at 10:34 am
Some other references on this topic include the workforce management work by K. Smilowitz (http://t.co/u4Q0nNlQSr) and our contribution on consistency – from the drivers’ point of view – in http://t.co/uzfnVf3GA3.
July 19th, 2013 at 11:25 am
Amazing two stories you have there, @Matforddavid