Waiting for a bus can be an exercise in patience, even when the buses are running on time. Contrary to what students might believe, IU Bus is very, very serious about making sure its service runs on time. In fact, with the help of the School of Informatics and Computing, it’s using the latest technology and the study of big data to try to improve its schedule and efficiency.
The effort is a collaboration between Mehmet Dalkilic, an associate professor of informatics and computing, and IU Campus Bus Service. Dalkilic was teaching a graduate-level course on data mining, and he was looking for real-world data that could be used to teach his class. He connected with Logan Keating, the Student Supervisor for IU Bus, who happened to be sitting on a large amount of data generated by their buses.
“In the fall of 2015, we had collected a lot of data, but we weren’t doing anything with it,” Keating said. “The main thing we wanted was variance. We wanted to know how far off our buses were from the schedules we posted. Memo’s class looked at the data, and some of it was favorable and some of it was not favorable. But it also turned out that variance wasn’t the best metric we could have used. I spoke with (IU Bus operations manager Perry Maull), who wanted us to continue the project. He was key to this project, and we all wanted to see what we could do to help the real-world performance of our buses.”
IU’s buses are equipped with a GPS that transmits data on the velocity, position, and heading of the bus every two seconds. There also are infrared sensors at each door, allowing the buses to track the number of passengers on the bus at any moment in time.
IU Bus hired two of Dalkilic’s data science students—Kurt Zimmer and Soumik Dey—on a part-time basis to perform data analysis and take the project to the next step—improving the schedule for the bus. A pilot program focused on the A-route Monday-Thursday, which runs from Memorial Stadium to the Wells Library to Third Street to Indiana Ave., and back again. The route is scheduled to take 27 minutes from start to finish.
“We found that it can actually take anywhere from 21 minutes to 35 minutes depending on traffic,” Keating said. “That’s a significant difference. If you can predict when the busiest times will come, you can build the schedule around them.”
That includes running more buses on the A route during the busiest times while not running quite as many when the traffic isn’t so heavy. The results are buses that run more efficiently and better scheduling of both routes and drivers.
“We’ve optimized for most trip times,” Keating said. “We’ve added nine percent more with the same man hours and the same layover percentage. Our drivers get the same amount of breaks with basically 10 percent more trips and the same number of hours of work.”
IU Bus is going to judge how well the system works during the fall semester of 2016 and make adjustments going forward. Dalkilic also hopes to expand and scale the system to Bloomington Transit with an eye on possibly taking the effort to other college campuses and public transit systems.
“This really is a wonderful example of a great collaboration here at IU,” Dalkilic said. “It’s a real-world example of the power of the work being done at SoIC, it shows the power of the process to bring together people when there’s a common need, and it shows the power of looking at big data in the large to help people make decisions with computation.”
Keating can envision an interactive process where municipal or private bus systems can use the scheduling process to either prioritize trips or passenger counts to optimize revenue. Whatever the use, Bloomington provided a unique opportunity to study the data.
“At IU, we have a perfect storm of me, who is a big data nerd, my boss, who is open to technology, and people here at SoIC,” Keating said. “It’s the perfect confluence of enough people who wanted to do this and were willing to commit to doing what needed to be done. I’m excited about it. I think it’s a promising project.”