A hybrid approach of genetic algorithms and local optimizers in cell loading

In this paper, the potential application of genetic algorithms to cell loading is discussed. The objective is to minimize the number of tardy jobs. Three different approaches are proposed and later compared. The first approach consists of two steps where (1) genetic algorithms is used to generate a job sequence and (2) a classical scheduling rule is used to assign jobs to the cells. The second approach consists of three steps where steps 1 and 2 are identical to the first approach plus step (3) Local Optimizer is applied to each cell independently. The third approach is very similar to the second approach except that chromosomes are modified to reflect the changes due to learning with local optimizer. Experimentation results show that the number of cells and the crossover strategy adapted affect the number of tardy jobs found. The results also indicate that hybrid GA-local optimizer approach improves the solution quality drastically. However, it has been also shown that GA alone can duplicate the performance of the hybrid approach with increased population size and number of generations in some of the cases. Finally, the impact of learning on the solution quality was not as significant as expected.