Forest-based Tabu Search to the Split-delivery Capacitated Arc-routing Problem

Abstract

Arc-routing problems (ARPs) aim to determine the least-cost vehicle routes of a specified arc subset in a graph. The split-delivery capacitated arc-routing problem (SDCARP) generalizes conventional ARPs by allowing an arc to be serviced by more than one vehicle. Because of its great complexity, the existing approaches for solving SDCARP are scant and most of them are meta-heuristics. In this paper, we analyze the characteristics of SDCARP solution structures. Then we devise a forest-based tabu search algorithm, which involves several forest-based neighborhood operators, for seeking the optimal solution. By using our method, the search space can be significantly reduced. The computational results demonstrate that our approach outperforms previous methods and is able to improve several state-of-the-art solutions on SDCARP benchmark instances.