Integrated GIS-based optimization of railway fuel tanker routing using Clarke-Wright initialization and local search

Abstract

Efficient fuel supply for special railway rolling stock requires a routing framework that combines realistic network data, fast initial construction, and practical improvement procedures. This paper presents a hybrid approach for railway fuel tanker routing in which GIS-based road distances are used to formulate a capacitated vehicle routing problem, the Clarke Wright savings algorithm generates the initial plan, and local search operators are applied as the main improvement and validation mechanism. The proposed framework is grounded in earlier work on integrated tanker optimization, simulation modeling, and GIS-supported railway logistics. A case study based on a West Siberian Railway section with one depot and four customer stations is used to demonstrate the method. The baseline solution of four separate routes totaling 381.6km is reduced to two feasible routes totaling 225.4km. This corresponds to a 40.9% reduction in route length while maintaining full utilization of tanker capacity on both routes. Local search confirms that the final route arrangement is locally optimal for the studied instance, which highlights its role as an essential post-optimization layer even when the constructive heuristic already performs well.