Rebalancing and Bikeshare

The mythical #DivvyRed

Is this the summer of Bikeshare? Divvy Bikes in Chicago launched last month. CitiBikes in New York City launched around Memorial Day. Any time now Bay Area Bike Share will be launching in San Francisco and on then Peninsula. 

The issue of having bikes where people want them is a perennial issue for bikeshare systems. "Rebalancing" is the act of moving inventory around to match demand and travel patterns. This map provides realtime visualizations of the demand of bikeshare systems around the world. Researchers are working on solving the rebalancing problem

A new article from EURO Journal on Transportation and Logistics works to develop a model for rebalancing. "Static repositioning in a bike-sharing system: models and solution approaches" by Tal Raviv, Michal Tzur, and Iris A. Forma, looks at how rebalancing or repositioning can help bikeshare systems.

Bike-sharing systems allow people to rent a bicycle at one of many automatic rental stations scattered around the city, use them for a short journey and return them at any station in the city. A crucial factor for the success of a bike-sharing system is its ability to meet the fluctuating demand for bicycles and for vacant lockers at each station. This is achieved by means of a repositioning operation, which consists of removing bicycles from some stations and transferring them to other stations, using a dedicated fleet of trucks. Operating such a fleet in a large bike-sharing system is an intricate problem consisting of decisions regarding the routes that the vehicles should follow and the number of bicycles that should be removed or placed at each station on each visit of the vehicles. In this paper, we present our modeling approach to the problem that generalizes existing routing models in the literature. This is done by introducing a unique convex objective function as well as time-related considerations. We present two mixed integer linear program formulations, discuss the assumptions associated with each, strengthen them by several valid inequalities and dominance rules, and compare their performances through an extensive numerical study. The results indicate that one of the formulations is very effective in obtaining high quality solutions to real life instances of the problem consisting of up to 104 stations and two vehicles. Finally, we draw insights on the characteristics of good solutions.

The full paper can be found here