Neural-Network-Based Nonlinear Optimal Terminal Guidance With Impact Angle Constraints

The terminal guidance problem considering nonlinearity, optimality, and impact angle constraints is investigated. First, the conditions for optimal guidance in the longitudinal plane are derived based on the Pontryagin's maximum principle, and then the to-be-solved two-point boundary value problem is equivalent to a backward integration problem. Then, analytical boundaries are given to initialize the states for backward integration. Based on the easily accessible dataset, a neural network is trained to approximate the optimal guidance commands. Lastly, an optimal terminal guidance scheme combined with the neural network and a biased proportional navigation guidance is proposed. Compared with the existing terminal guidance methods, the proposed guidance strategy balances the performances about flight optimality, on-board implementation capability, and impact angle satisfaction when high dynamical nonlinearity is considered. Simulations are given to validate the effectiveness of the proposed techniques, and demonstrate the advantages of the algorithm on optimality, real-time performance, and impact angle satisfaction in nonlinear cases.