Evaluation of Computational Methods for Freight Train Classification

Abstract

In this paper we study a special sorting problem from real-world railways called train classification. Applying an efficient encoding of classification schedules developed in our previous publication, we derive an integer programming model for deriving such schedules which is capable of incorporating many different practical restrictions easily. We evaluate our model extensively using various synthetical instances and apply it to a real-world traffic instance to obtain a schedule that improves on the method currently used there.