Abstract: The traditional genetic algorithm for solving flexible job shop scheduling problem tends to fall into slow convergence and local optimum.Aiming at minimizing the largest makespan for flexible job shop scheduling problem,an improved multi-population genetic algorithm (IMGA) is proposed in this paper.The lateral and longitudinal evolutionary mechanism is used.The directed evolution is used for genetic algorithm,and directed evolution probability is also designed in order to accelerate the speed for getting the optimal solution.In the selection process,the resurrection strategy is used to avoid losing some good genes by combining eliminated individual genes with outstanding individual genes during the evolving operation.The experimental results demonstrated that the proposed IMGA is feasible and effective.