An intensification approach based on fitness landscape characteristics for job shop scheduling problem

Abstract

This work deals with the classical Job Shop Scheduling Problem (JSSP) of minimizing the makespan. Metaheuristics are often used on the JSSP solution, but a performance comparable to the state-of-the-art depends on an efficient exploration of the solutions space characteristics. Thus, it is proposed an intensification approach based on the concepts of attraction basins and big valley. Suboptimal solutions obtained by the metaheuristic genetic algorithm are selected and subjected to intensification, in which a binary Bidimensional Genetic Algorithm (BGA) is utilized to enlarge the search neighborhood from a current solution, to escape of attraction basins. Then, the best solution found in this neighborhood is used as the final point of the path relinking strategy derived from the initial suboptimal solution, for exploring possible big valleys. Finally, the best solution in the path is inserted into the population. Trials with usual instances of the literature show that the proposed approach yields greater results with regards to local search, based on permutation of operations on critical blocks, either on the makespan reduction or on the number of generations, and competitive results regarding the contemporary literature.