An efficient discrete invasive weed optimization for blocking flow-shop scheduling problem

This paper proposes a discrete invasive weed optimization (DIWO) to solve the blocking flow-shop scheduling problem (BFSP) with makespan criterion, which has important practical applications in modern industry. In the proposed DIWO, an effective heuristic and the random method are combined to generate an initial plant population with high quality and diversity. To keep the searching ability and efficiency, a random-insertion-based spatial dispersal is presented by means of the normal distribution. Moreover, a shuffle-based referenced local search is embedded to further enhance local exploitation ability. An improved competitive exclusion is developed to determine an offspring plant population with good quality and diversity. The parameters setting is investigated based on a design-of-experiment approach. The effectiveness and applicability of the proposed spatial dispersal and local search are confirmed through numerical comparisons. Finally, a comprehensive computational evaluation including several state-of-the-art algorithms, together with statistical analyses, show that the proposed DIWO algorithm produces better results than all compared algorithms by significant margin.