New tool to help Metro Transit schedule drivers, optimize workforce planning
A U of M team has completed a project aimed at helping Metro Transit schedule and manage its bus operator workforce. Led by Qie He, an assistant professor in the Department of Industrial and Systems Engineering (ISyE), the project was part of a multi-year partnership between the University and the Twin Cities transit agency.
Metro Transit’s extensive network of bus service employs more than 1,500 bus operators at five garages. Every day, operations staff must manage this workforce, adjusting for both planned and unplanned absences. This open work can be assigned to a reserve operator (who is paid whether assigned work or not) or to a regular operator who is asked to work overtime. “Our project developed an analytics tool to help the agency minimize service disruption while also minimizing total costs,” He says.
During the study, ISyE students were immersed at Metro Transit: they worked in garages, met with dispatchers, and connected with workforce planning staff on a weekly basis. “The research team did a great job learning about our operations, the constraints that exist, and the day-to-day challenges we face,” says Donathan Brown, assistant director of bus operations administration for Metro Transit. “They then applied their analytical skills and gave us important insight into our operations.”
With the insight gained from these interactions, researchers created a machine-learning model that predicts how many drivers will be absent at a given point in the future. “This tool can predict absences of different types at the individual level, the garage level, and across the entire organization,” He explains. “This model helps Metro Transit get a better understanding of the factors that affect operator absences as well as provides a prediction of the distribution of daily absences.”
The output of this prediction tool then feeds into another tool the research team created—an optimization model—that recommends how many reserve operators are needed for the next day. The goal is to balance the number of reserves with the number of regular operators who are asked to work overtime.
A key finding from the project is that the current number of assigned reserve drivers is slightly more than is needed for the entire organization. “Operator shortage may occur at only some garages during certain times,” He says, “so one approach is to rebalance reserves more frequently across garages.”
One next step in the work is to cross-check and validate the models with recent data. The team also hopes to develop and evaluate tools that can be used on a daily basis by operations staff to guide decisions on operator assignments for the next day.
“This is a great example of how theoretical research can be translated into improving our daily work,” says Eric Lind, manager of research and analytics at Metro Transit. “Ultimately, we will measure the real outcomes of this research in the effects on reducing both service disruptions and the cost to deliver service.”
Professor He also notes unmet research needs: “The most essential one is that the current optimization model cannot provide real-time assignment of reserve drivers,” He says. “Additional work could develop a real-time algorithm that assigns operators to a specific piece of work.”
The project was part of an external stakeholder engagement program launched in 2016 by the U of M’s Office of the Vice President for Research. Metro Transit cosponsored the project.