Large-scale crew scheduling method by switching allocation knowledge

In railway crew scheduling, it is a problem to generate a schedule of crew according to the train timetable, satisfying constraints such as working hours, recess hours, etc. Our research group has proposed a scheduling method which contains two phases: an initial scheduling phase using the state selection method and a modification phase using the relaxation search method. Though this method has been successfully applied to a real small-scale timetable, the initial scheduling is difficult for a large-scale timetable, because the feature of allocating trains changes during the planning process. In order to satisfy constraints of "meal recess" and "allocation of all trains", which are considered as especially difficult constraints, a method for the initial scheduling of a large-scale timetable is proposed. In this method, the planning features change is coped with by switching allocation knowledge. The driving rate is introduced to decide the point of changing knowledge. Four types of knowledge are created to satisfy constraints of "recess for a meal" and "allocation of all trains" for each planning feature. This method is applied to a real large-scale timetable. It is confirmed that the method can make up the initial schedule which satisfies the constraints of "meal recess" and "allocation of all trains".