Working for Public Works
When two large Twin Cities public entities called, ISyE assistant professor Qie He was happy to offer his expertise. “I like theory, but I consider myself an engineer first. I love to solve real-world problems.”
Predicting Driver Absences to Minimize Service Disruption
The first project came from Metro Transit, the organization which manages nearly all Twin Cities public transportation – to the tune of 1000 buses and 1500 drivers that cover 130 daily routes over a 907 square mile area. Metro Transit’s problem: frequent and uncertain driver absences.
Metro Transit has full-time reserve drivers available daily, but seemingly random spikes in absences meant that overtime had to be used for routes on many days, leading to additional costs. Metro Transit, along with the University’s Office of Vice President for Research, co-sponsored the work of Qie He, graduate student Xiaochen Zhang, and ISyE senior Soniya Somani.
He and his team compiled absence and overtime data from previous years and started to see patterns. There were more absences during the school year, for example, and early morning routes needed reserves more often than daytime routes. The team created a machine-learning model to predict driver absences each day and an optimization model to recommend the number of reserves for the next day, which they then validated using more recent data.
“This is a great example of how theoretical research can improve our daily work,” says Eric Lind, manager of research and analytics at Metro Transit.
Optimizing Maintenance and Purchasing
Qie He’s second project came from Hennepin County – Minnesota’s largest county by population – and was a collaboration with fellow ISyE assistant professor Dan Mitchell. The county asked Mitchell and He to review its system for managing its fleet of over 250 pieces of mobile equipment. In particular, the county didn’t know if its policy of retiring snowplows after 10 years was optimal.
He and Mitchell focused on the county’s tandems, the trucks that are used for snowplowing and other regular duties. They developed models to predict the depreciation and maintenance costs of a truck based on 10 years’ historical data such as initial price, maintenance costs, resale price, and rental income. Then, they experimented with different purchase and resale policies to determine the best average cost per truck per year of service.
“Our program suggests that replacing these trucks every 11 years will result in a 3 percent savings in annual operating costs for the county,” says He, which would mean $2 million in savings over 10 years.
He and ISyE head of department Saif Benjaafar hope to engage with more municipalities. “It should be part of the department’s lifeblood,” says Benjaafar. “Municipal problems are fundamentally no different from the problems with which our industry partners present us every day. We can fulfill a real need, and students and faculty get to be engaged in real-world projects. It’s a win for everyone.”