A genetic algorithm with MGG and demand crossover to solve dynamic flexible scheduling problem

A genetic algorithm, called MDGA, is proposed for practical scheduling, where bills of materials of parts, routings of production operations, and work-in-process inventories on hand and in near future, are taken into consideration. The scheduling problem is called a dynamic flexible scheduling (DFS) problem. The MDGA algorithm uses the concept of basket of requirements in representation of chromosome. MDGA reproduces a population of chromosomes with the principle of minimum generation gap (Yamamura et al., 1996) instead of simple tournament selection in usual genetic algorithm. In order to demonstrate the correctness of MDGA, a comparison with exhaustive search is provided, which also shows the difficulty in solving the DFS problem. By applying MDGA to a usual job shop scheduling problem, which is a simplified DFS problem, the effectiveness of MDGA is shown to be satisfactory. Finally, since MDGA has many parameters, it is examined how they effect on solution-search process.