A metaheuristic algorithm for a locomotive routing problem arising in the steel industry

In this study, we address the locomotive routing problem faced in the steel industry. During steel production, locomotives are employed to move torpedo cars, which transport molten iron, between blast furnaces and converters. The resulting complex pickup and delivery routing system which features multiple types of practical constraints, such as hard time windows, incompatibility of requests, variable locomotives capacity, and last-in-first-out constraints, constitutes this problem. Herein, three-index and modified group-based formulations are described and an adaptive large neighborhood search (ALNS) algorithm is proposed as a solution. The ALNS algorithm relies on both adapted and novel removal and insertion operators and a feature-based local search procedure. The mathematical formulations and the ALNS algorithm are evaluated on instances generated according to the actual production process. The results validate the effectiveness of the algorithm. A comparison with the manual plan also verifies the quality of the solutions produced by the algorithm.