Incorporating Highly Explorative Methods to Improve the Performance of Variable Neighborhood Search

Variable Neighborhood Search (VNS) is one of the most recently introduced metaheuristics. Although VNS is successfully applied on various problem domains, there is still some room for it to get improved. While VNS has an efficient exploitation strategy, it suffers from its inefficient solution space exploration. To overcome this limitation, VNS can be joined with explorative methods such as Evolutionary Algorithms (EAs) which are global population-based search methods. Due to its effective search space exploration, Differential Evolution (DE) is a popular EA which is a great candidate to be joined with VNS. In this article, two different DEs are proposed to be combined with VNS. The first DE uses explorative evolutionary operators and the second one is a Multi-Population Differential Evolution (MP-DE). Incorporating a number of sub-populations improves the population diversity and increases the chance of reaching to unexplored regions. Both proposed hybrid methods are evaluated on the classical Job Shop Scheduling Problems. The experimental results reveal that the combination of VNS with more explorative method is more reliable to find acceptable solutions. Furthermore, the proposed methods offer competitive solutions compared to the state-of-the-art hybrid EAs proposed to solve JSSPs.