With an integrated flight/propulsion control system of a short take-off and landing fighter as design plant,an optimization design approach for decentralized control was proposed in order that it can maintain the performance and robustness of its corresponding centralized control.A centralized controller was firstly designed by using genetic algorithm as the performance reference for the decentralized controllers.The optimization design method for the decentralized controllers with interface variable was then provided.A critical element affecting decentralized control performance was analyzed and the decentralized control performance expression using frequency weighted matrix as design parameter was deduced.The problem of designing the decentralized controllers was converted into optimization problem solved by genetic algorithm.An improved balanced reduction method was also used in decentralized controller design.The simulation and structured singular value analysis show that the proposed optimization design method is so effective that the decentralized control system approximately reaches the same performance and robustness achieved by centralized control system.