Every day across the US, fleets of iconic yellow school buses carry more than 25 million children to and from school, covering around 6 billion kilometres a year. But these instantly recognizable vehicles - and the authorities that run them - are facing a number of challenges.
One major issue has been a shortage of drivers. From Hawaii to Nebraska, many school districts across the US have had to scale down or even suspend bus services. Even with signing-on bonuses and guaranteed minimum earnings, the town of Lincoln, Nebraska, couldn’t recruit enough drivers to fill all its vacancies.
The bus services are also expensive to run – fuel and maintenance costs are a major factor. Plus there is growing pressure to replace older, more polluting buses with newer models, which involves additional outlays.
A digital solution
But in Boston, automation may offer a route to solving some of these problems. Not in the form of self-driving vehicles, but in the use of machine-learning to improve school bus administration.
In 2017, Boston Public Schools announced a competition - the BPS Transportation Challenge - to find smart ways to improve school bus operations and bring down their $120 million annual cost.
The winner was the Quantum Team from the MIT Operations Research Center. The team developed an algorithm to identify the most efficient and cost-effective routes for BPS’s fleet of 650 buses. And last year, the algorithm got to work contending with multiple variables to reconfigure routes.
BPS gives parents a wide range of school choices, which means some schools have students from more than 20 different postal codes across Boston. Many schools in the district also operate different opening and closing hours, with start times varying between 07:15 and 09:30, forcing buses to make multiple trips with few students.
Next stop, progress
But wasteful journeys like those are becoming a thing of the past in Boston.
The MIT algorithm takes data from Google Maps to analyze traffic patterns. It combines that with details on where students live and the schools they attend to calculate the best possible routes.
Huge amounts of time have also been saved for teams of school bus planners, as it takes the algorithm just 30 minutes to do a job that used to take several weeks to complete manually.
Last year, BPS buses drove 1.6 million fewer kilometres, reducing daily CO2 emissions by 9,000 kilograms. It also plans to use the algorithm for an overhaul of school start and finish times, as part of the district's ongoing initiative to increase efficiency and effectiveness.