Train routing in shunting yards using Answer Set Programming

Abstract

We use Answer Set Programming (ASP), a modern method for knowledge representation and reasoning, to build a fully automated decision support system for routing of trains in shunting yards. The system leverages the knowledge of experienced shunting yard operators and yields optimal, consistent and transparent routing decisions. In addition, the system remains easily adaptable to new expert knowledge that may become available in the future. We embedded this routing system into a simulation environment and conducted a study in order to investigate and confirm the validity and limits of this new approach. The study is based on the track layout and legal regularities of an actual shunting yard and therefore ensures the applicability to real world problem instances. The results confirm that ASP can be used to solve complex routing problems, but cannot yet match the solving speed of proprietary and custom fit algorithms. Therefore the suitability of ASP to solve complex routing problems is subject to the trade-off between transparency, adaptability and flexibility vs. speed.