In this paper, we propose a new Genetic Algorithm for JSP using two crossovers. The crossover, JOX, obtained relatively good results, however offspring generated by JOX exist around parents or within an intermediate area of them. This feature of JOX induces a convergence of the whole population. To deal with this fault of JOX, we propose a complementary combination of two crossovers. One is JOX, and the other, EDX, is our proposal. EDX is designed to have the population enlarge using a local search and explores the area where the population uncovers. Although a mutation is applied for exploration in general, we apply a framework of crossover to EDX for a more efficient exploration. The combination of two crossovers, which has a different search area, is able to compensate for each other's fault. The GA designed with these two crossovers was applied to large-size JSP benchmarks, and we show its effectiveness.
展开▼