A Genetic Algorithm for the mass transit crew rostering problem

Abstract

In this paper the Crew Rostering Problem (CRP) from a Brazilian public transit company is modeled and solved by a Genetic Algorithm (GA). The CRP consists of assigning duties on weekdays, Saturdays and Sundays to the crew members of a company over a given planning horizon, minimizing the number of crews and also balancing the workload of the crews, reducing the total number of accumulated overtime and idle time for each crew, subject to a set of operational and legal constraints. The CRP is solved in two stages: defining the sequence of working and rest periods, and defining the daily assignment to be performed in each working period by each employee. The GA was tested with data from a medium sized Brazilian company and was shown to be efficient.