A Neural Net Approach to Space Vehicle Guidance

Abstract

The on-line implementation of numerical algorithms for solving the optimum trajectory/guidance problem for advanced space vehicles such as ALS, HLLV, AOTV, transatmospheric vehicles and interplanetary spacecraft is not possible due to their complexity. Hence, the current approach to the development of real-time guidance laws for these advanced space vehicles is to use approximation theory to obtain closed-loop guidance laws. Neural networks offer an alternative to the derivation and implementation of guidance laws. In this paper, we formulate the space vehicle guidance problem using a neural network approach and investigate the appropriate neural net architecture for modelling optimum guidance trajectories. In particular, we investigate the incorporation of a priori knowledge about the characteristics of the optimal guidance solution into the neural network architecture. The online classification performance of the developed network is demonstrated using a synthesized network trained with a data base of optimum guidance trajectories. Such a neural network based guidance approach can readily adapt to environment uncertainties such as those encountered by an AOTV during atmospheric maneuvers.