An Improved LP-Based Heuristic for Solving a Real-World Locomotive Assignment Problem

Abstract

The locomotive assignment (or scheduling) problem is a highly relevant problem in rail freight transport. For a preplanned train schedule, minimum-cost locomotive schedules have to be created so that each train is pulled by the required number of locomotives (locomotives are assigned to trains). Determining locomotive schedules goes hand in hand with determining the number of required locomotives and this has a significant impact on capital commitment costs. Therefore, this paper proposes an improved heuristic for scheduling locomotives at a European rail freight operator. We show that a transformation of an iterative process to simplify the underlying network into a one-step procedure can significantly reduce computing times of a heuristic. Computational tests are carried out on the real-world instance as well as on smaller instances. The results show that the proposed heuristic outperforms an existing heuristic from literature in terms of both solution quality and computation times and, in contrast to approaches from literature, enables a solution of a practical instance in Europe.