Building a Smarter (and Cheaper) School Bus System: How a Boston-MIT Partnership Led to New Routes That Are 20% More Efficient and Saved the District $5 Million
- .@MIT computer scientists rework Boston school bus routes to cut number of vehicles and miles driven — & save district million of dollars
- Unique partnership with @MIT Operations Research Center gave Boston district school bus routes that were 20% more efficient, used 400 fewer buses and saved at least $5 million
Updated August 26
Boston Public Schools had a problem. The district needed to figure out the best way to route its school buses, but solutions already on the market weren’t working. It couldn’t do the job on its own. So the district launched a public competition to get outside help.
The winner, a team from MIT’s Operations Research Center, created an algorithm that saved Boston schools $5 million and 1 million miles driven in its first year. One of the most expensive school transportation departments in the country was able to reduce its fleet by 8 percent, the largest single year-over-year drop in district history. And today, two years in, the district has dropped its number of trips by nearly 400, to 2,800, using around 620 buses in a system that is 20 percent more efficient.
“They were the clear winner,” says Will Eger, Boston Public Schools’ strategic projects manager — and that winning solution, which did in just 30 minutes what used to take a team of 10 staffers 3,000 hours to accomplish, can help districts all across the nation.
Invest in independent journalism. And help The 74 make an impact.
Help us reach our Spring Campaign membership goal.
“I think the reason routing school buses is so hard is because essentially every year you have to start from scratch as kids move grades, switch schools, and enter and leave the system,” says Arthur Delarue, an MIT Ph.D. student who worked on the project with Sebastien Martin and Dimitris Bertsimas. “Essentially, the work you did to route the system the previous year is just lost. You have to restart.”
While the best routing algorithms work well for logistics companies like FedEx and UPS, they don’t apply to school districts with specific constraints dictated by community priorities, state regulations and the local environment. “It is hard to have a framework that generalizes all these different types of problems,” Delarue says. “Every school district has a problem that is slightly different. This is just a mathematically difficult problem, one of the most difficult problems you can think of. The number of possibilities is enormous, and you need to prune all the bad solutions quickly to find good solutions.”
What makes Boston particularly difficult is that the 25,000 students who use the district’s transportation service stretch across 20 zip codes and have the option of up to 10 schools out of more than 220 in the district. While some districts simply create fixed routes no matter where students live, Boston’s policy is to provide service for anyone living more than a half-mile from a school and limit the time students spend on the bus, all while placing stops at the closest possible intersections. “For us, this is constantly changing,” Eger says. “That level of change requires moving bus stops, changing routes, which requires a pretty complicated level of agility. We want to be as minimally disruptive to student rides as much as possible, but we also want to be responsive to changes in enrollment.”
The best solution turned out to be not the one with the fewest buses per each school, but one that reuses buses from school to school.
“This problem is so hard, there is no way to find the perfect solution,” says Martin, also a Ph.D. candidate at MIT. “We identified what was best for one school, but if you do this for each school, then there is actually a big problem in the system. What is best for one school is not always best for the way schools interact with each other. If you don’t take into account the need to reuse buses, you will need a lot of buses and cost a lot to the district. We consider several different solutions for each school and look at what interfaces best overall.”
Most buses are shared by three schools in the morning and again in the afternoon, taking into account the various types of buses and their differing capacities based on students’ ages.
“Routing more efficiently is hugely important, and reinvesting in classrooms is an amazingly huge win from this project, but also the ability to explore options with really good levels of accuracy is really exciting,” Eger says. Now, the district can weigh how potential changes in policy may impact costs. If the district, for example, looks at providing bus service only to students living three-quarters of a mile from a school, it can use the MIT solution to understand the cost. “It opens up a whole new world for us to be much smarter as we explore policy trade-offs,” he says.
The ongoing work the team is doing with Boston has led Bertsimas, an MIT professor, to launch Dynamic Ideas, a company marketing the product to other school districts, now in the early stages of adoption.
“It is wonderful other school districts can benefit in the work we helped create,” Eger says. Martin says he enjoys seeing research move to deployment, because researchers often strip away excess problems to get to the core of an issue; but in order to make something useful for schools, the MIT team had to add in all sorts of real-world complexities. It was a mixture of math and human involvement.
“There are many things you can’t anticipate, and there is a team of roughly 10 transportation officers [in Boston] who do a fantastic job to work on solutions and modify when unexpected things come up,” Martin says. “It is still a human process.”
As Dynamic Ideas ramps up, its team knows that every district has a slightly different problem, and that although the algorithm can be applied in a wide variety of settings, it takes time to adjust and solve multiple problems.
As for Boston, having seen what this collaboration could do, Eger is excited about what other wins he can achieve by bringing in outside help. “I feel like this project has opened up my eyes to the possibility of partnering with incredibly talented researchers in places folks don’t often look,” he says. “I didn’t know about the MIT Research and Operations Center, and I’m excited to think about what other problems we can work with them on for huge impacts.”
Says Delarue, “Boston Public Schools was exceptional in their ability to recognize that this is something they should address with external solutions. They have driven this conversation forward across the country.”Submit a Letter to the Editor