基于分布式并行伪谱-神经网络算法的双脉冲导弹多阶段协同轨迹优化

A decentralized parallel algorithm based on Gauss pseudospectral method and artificial neural network is presented for the multi-stage cooperative trajectory optimization problem of dual-pulse missiles, in which considers dropping the detached engine shell to a safe area. According to the characteristic of two-stage pulse, the entire trajectory of dual-pulse missile is divided into four flight stages, such as launching, climb, endurance and attack. In order to predict the landing position of the detached engine shell, a range prediction function is established and is fitted off-line by artificial neural network to improve the prediction speed. In the decentralized parallel algorithm, each missile independently solves its optimal trajectory in parallel. The lower bound constraint of flight time is introduced to ensure the flight time consistency of missiles. The decentralized parallel algorithm is compared with the centralized algorithm by two simulation examples. The simulated results show that the decentralized parallel algorithm proposed in this paper can be used to obtain better performance index and higher solution efficiency for solving the multi-stage cooperative trajectory optimization problem of dual-pulse missiles.