The approximation capability of artificial neural networks has been applied to the midcourse guidance problem to overcome the difficulty of deriving an on-board guidance algorithm based on optimal control theory. This approach is to train a neural network to approximate the optimal guidance law in feedback form using the optimal trajectories computed in advance, Then the trained network is suitable for real-Lime implementation as well as generating suboptimal commands. In this paper, the advancement of the neural-network approach to the current level from the design procedure to the three-dimensional flight is described. (C) 2001 Published by Elsevier Science Ltd.