Crew Decision Assist: System for Optimizing Crew Assignments at BNSF Railway

Rail is the preferred mode of transport for many categories of freight because of its low cost and energy efficiency. Rail accounts for approximately 40%, measured in ton-miles, of all freight movements in the United States. To maintain their competitive advantage and effectively utilize their large investments in rail infrastructure, freight railroad companies place considerable emphasis on improving the cost efficiency of their operations. Crew costs, including payments to crew members and expenses for crew repositioning and lodging at stations away from the home base, constitute a significant portion of railroad operating expenses. This paper describes the development of an optimization model and solution method and the implementation of a system called “crew decision assist” to support crew scheduling at BNSF Railway. The work was motivated by the company’s desire to replace its current manual crew-planning process with a systematic and effective approach. Preexisting crew-scheduling models did not adequately capture all the options and constraints that arise in practice, such as the option to use extra crew members or policies to jointly reposition engineers and conductors. We, therefore, developed a tailored model and solution approach that incorporates various practical features and requirements for crew assignment at BNSF and accounts for uncertainty in train schedules. Our decision support system, based on this method, interfaces with existing information systems to retrieve the necessary data and quickly generate effective crew-deployment plans when train schedules change. The system was recently introduced for use by crew planners at BNSF and has already reduced crew costs, yielding estimated annual savings of several million dollars.