Optimal management of full train load services in the shunting yard: A comprehensive study on Shunt-In Shunt-Out policies

A key aim of the European Union is to double freight rail traffic by 2050 in order to reduce pollution emissions and alleviate congestion by shifting traffic from roads to rail networks. To accomplish this objective, it is crucial to minimize emissions and high costs associated with shunting yard operations while maintaining an acceptable level of service. This research paper introduces a new MINLP model that optimizes the shunt-in and shunt-out (SISO) operations of wagons in a shunting yard that handles full train load services. Additionally, an efficient Multi-Objective Dijkstra Algorithm (MDA) is proposed to handle simultaneous shunt-out operations in a multi-train scenario. The MINLP model takes a mesoscopic approach, aiming to minimize the number of SISO operations while satisfying strategic and tactical objectives such as wagon fleet size, operational costs, and shunting locomotive emissions. Several versions of the mathematical model are described, each employing different Shunt-In (SI) policies with varying criteria for wagon selection and strong goal orientation. The Multi-Objective Dijkstra Algorithm determines the Multi-Objective Shortest Path between mandatory shunts, considering both the time required for shunting and clustering costs. It provides information on the clusters of wagons that need to be shunted out. To assess the effectiveness of the MINLP model, real train timetables for freight trains in the Bettembourg Eurohub Sud Terminal (Luxembourg) are used, and various KPIs related to tactical and strategic objectives are evaluated. Furthermore, the performance of the Multi-Objective Dijkstra Algorithm is compared to the Shunt-Out sub-model, considering average computation time and solution quality in relation to the MINLP model. The computational results demonstrate that the criteria for wagon selection have a significant impact on the analyzed KPIs.