Trajectory Optimization of Hypersonic Glide Vehicle Based on Pseudospectral Method Assisted by BP Neural Network

Pseudospectral method is often used to find the optimal trajectory of hypersonic glide vehicle. Considering the influence of the initial guess value on the calculation speed of the pseudospectral method, this paper presents a trajectory optimization strategy using BP neural network to generate the initial value of pseudospectral collocation points to assist the calculation of pseudospectral method. The dataset is obtained by offline optimization using pseudospectral method under random flight missions. By using specific flight mission generalization methods, the neural network scale is reduced and the convergence speed of the network is improved. A BP neural network model is established with the generalized flight mission as input and the initial guess value of pseudospectral collocation points as output. Numerical simulations in validation set show that, compared with the pseudospectral method based on the traditional initial value generation method, pseudospectral method assisted by neural network has significantly improved in solving speed and reliability.