Scheduling multi-objective job shops using a memetic algorithm based on differential evolution

In this paper, a memetic algorithm based on differential evolution (DE), named MODEMA, is proposed for multi-objective job shop scheduling problems (MJSSPs). To balance the exploration and exploitation abilities, both DE-based global search and an adaptive local search are designed and applied simultaneously in the proposed MODEMA. Firstly, a smallest-order-value (SOV) rule is presented to convert the continuous values of individuals (real vectors) in DE to job permutations. Secondly, after the exploration based on DE, several neighborhoods are used in a local search and an adaptive Meta-Lamarckian strategy is employed to dynamically decide which neighborhood should be selected to stress exploitation in each generation. In addition, a solution set is used in MODEMA to hold and update the obtained nondominated solutions. Simulation results and comparisons with Ishibuchi and Murata's multi-objective genetic local search (IMMOGLS) show the effectiveness and robustness of the proposed MODEMA.