Carnegie Mellon University

How To Keep Inventory Balanced in the Bike Sharing Economy

Willem-Jan van Hoeve, Carnegie Bosch Associate Professor of Operations Research, speaks about his research on inventory rebalancing and vehicle routing in bike sharing systems.

Video Transcript

You have a business problem, and you translate that into a mathematical model. That is, I think, the most important part of operations research. And one paper, in particular, is about bike sharing systems. You have to make sure when a user wants to use a bike, it should be there, and when you want to drop off the bike there should be an empty spot.

The issue we have is that, of course, during commuting hours people will go from one place to their work environment and then they want to go back in the afternoon when they commute, and so you get piles of bikes in one part of the day and one area of town and, of course, there's a lot of demand of bikes that you cannot deliver, perhaps. So you get an imbalance in these bike stations and the inventory in terms of bikes, and also in terms of parking spaces. And we have a predictive model, a mathematical model that describes at each part of the day, say 8 a.m. or 9 a.m., how many people want to get a bike from the station and how many people want to drop off their bike at the station. And now it's our role as an optimization or an operations researcher to make sure that we redistribute the bikes over time so that we have enough of this demand fulfilled.

So we have to design these routes for these trucks to pick up a certain amount here, deliver it there, and so we use the predictive model to get a lower and upper bound, then intervalling what is the level of inventory at each of these stations, and we then optimize the route of trucks to make sure that we are within those bounds and we can meet as much demand as we can in the, say, coming 2 or 3 hours. It's a very challenging problem but we found a way to decompose it into independent smaller clusters. It can be rebalanced independently and it scales dramatically better.