A big problem for any bike sharing program is the “rebalancing” of the system, or finding a way to keep a good number of bikes at all the stations. In an effort to address this problem in the local program, Houston Bcycle collaborated with the Oshman Engineering Design Kitchen at Rice University. We (Latane Bullock, Patrick Han, Sam Robedee, and David Senter) are undergraduate engineering students at Rice known as “The B Team”. Over the course of a semester, we developed a computer algorithm that tells Bcycle's field technicians how to move bikes to improve the distribution of bikes in the system.
From L to R: Patrick Han, Latané Bullock, David Senter, Sam Robedee
The algorithm, which runs several times a day, works in three steps to produce simple instructions for the Bcycle field techs. First, it looks at historical data that shows how many bikes are at each of the stations and the trips that were taken throughout the day. Next, it identifies patterns in the data and labels which stations usually need more bikes and which stations tend to have extra bikes. In the final step, the algorithm pairs nearby stations from each of these groups. These pairs of stations are sent to the field techs in order of importance. An example of one of the instructions might be, “Take 4 bikes from Sabine Bridge to Brazos & McGowen.” Filed techs can then decide if the instructions make sense and move the bikes accordingly. Ideally, the instructions can help keep stations from being too empty or too full, especially as Bcycle adds more stations to the system in the next few months.
Designing the algorithm was a rewarding experience overall, but it was certainly challenging at times. For example, we initially had trouble getting the program to read and sort the tens of thousands of lines of bicycle trip data. A single extra comma could cause a critical error in the algorithm’s calculations. Although we encountered a number of obstacles like this and made several difficult decisions throughout the design process, we overcame them and produced something we are proud of. It was an exceptional opportunity for us to learn how to work on a project as a team with the community in mind.
From the results of our initial testing, we recorded a noticeable improvement in the system’s balance when compared to the past. While our limited testing data may not be sufficient to declare our solution as completely effective, the feedback we received from the field techs confirmed that, at the very least, our solution acted as a useful aid. As the system triples in size, the field techs may find it more difficult to rely on their intuition and routine alone. We believe our algorithm will be even more helpful as the system expands.
We would like to thank Houston Bike Share executive director Carter Stern as well as all our instructors at the Oshman Engineering Design Kitchen at Rice University for giving us the opportunity and support we needed to address the rebalancing problem.
“The B Team”
Latane Bullock, Patrick Han, Sam Robedee, and David Senter