An improved island model memetic algorithm with a new cooperation phase for multi-objective job shop scheduling problem

- A new naturally inspired cooperation phase.
- Two types of knowledge: already existed and recently acquired one.
- Knowledge is exchanged in favor of the recently acquired with a fixed mixing ratio.
- Tested on 72 benchmark instances, and compared with other similar works.
- Computational results validate the effectiveness and robustness of the proposed algorithm.

This work proposes an improved island model memetic algorithm with a new naturally inspired cooperation phase (IIMMA) for multi-objective job shop scheduling problem. Three objective functions: makespan, total weighted tardiness, and total weighted earliness are considered using the weighting approach. The new cooperation phase is mainly used to improve the exploitation capabilities of an island model memetic algorithm. It is based on the following novel idea. Individuals who have recently performed self-adaptation phases (local search) do not exchange their knowledge about the search space just randomly; instead, they firstly divide their current knowledge into two parts: already existed knowledge and recently acquired knowledge, and secondly exchange their knowledge in favor of the recently acquired one. This is simulated by means of an improved version of the well-known uniform crossover, which uses the history of parents' evolution to identify the new traits among the old ones, and then to construct the mask vectors that determine the exchanged genetic materials accordingly. Additionally, several straightforward but effective techniques are applied to improve the exploration capabilities as well, such as a diversity-based population creation method, an incest prevention-based tournament selection method, and a similarity-and-quality based replacement method. The presented algorithm is evaluated on 72 benchmarks, with the new components, and without them using the traditional alternatives, and also against similar works found in the literature. The computational results validate the improvements accomplished by the new components, and show its effectiveness and robustness.