Cooperative coevolutionary algorithm for job shop scheduling problem

An improved cooperative coevolutionary algorithm was proposed to solve job shop scheduling problem. According to the number of machines, the whole population was naturally divided into some subpopulation whose individuals encoded the preference list of jobs on the corresponding machine. The steady-state reproduction was introduced to genetic operators. The proposed algorithm combined three types of cooperative partners from every other subpopulation with the evaluated individuals to form the whole solutions and adopted the improved preference-list-based G and T algorithm to decode them to evaluate. Finally an updating technology and dynamic substitution with some new individuals at some other generations was adopted to speed up the convergence. Numerical experiments have been made to solve some job shop benchmark problems. The optimization results show the proposed algorithm have outperformed traditional genetic algorithms.