Fare inspection optimization in train networks

Abstract

In this thesis we present the inspector scheduling problem set in the local train network around Oslo serviced by Norges Statsbaner (NSB). We review current literature on the subject and present our own solution method adapted to NSB's inspection policy. By using mathematical optimization and specifically column generation we construct an optimal set of patrols plans and a corresponding probability distribution. Inspections are randomized by sampling patrols each work day. Solving the optimization problem presents computational challenges and we present a heuristic method for finding approximate solutions. The solution method is validated through several numerical experiments using example train networks inspired by the actual NSB local train network. The results suggest that our method is viable for practical applications although some work remains before this can be realized.