Adaptive Autopilot Design for Guided Munitions

Traditional autopilot design for guided munitions requires an accurate aerodynamic model and relies on a gain schedule to account for system nonlinearities. This paper presents an approach that simplifies the autopilot design process by mating an inverting controller designed at a single flight condition with an on-line neural network to account for errors that arise due to the approximate inversion. This eliminates the need for an extensive design process and also the requirement for accurate aerodynamic data, which can be especially critical at high angles of attack or other regimes at which the aerodynamics become highly nonlinear. The choice of inversion process itself has been found to be critical in the implementation, and is therefore discussed at length. Finally, results from an application of this approach to a full nonlinear 6DOF guided munition simulation are presented.