Application of an Adaptive Autopilot Design to a Family of Guided Munitions

A previously developed autopilot featuring a dynamic inversion control law augmented by an on-line neural network is applied to control a variant of the Joint Direct Attack Munition (JDAM). The inverting control law is based on a simple linear model of the plant at a single flight condition. The neural network adaptively regulates error in the plant inversion that occurs throughout the flight envelope due to the use of an approximate model. The autopilot design is performed using minimal information from an approximate aerodynamic data set generated using Missile DATCOM. The design is then validated in a nonlinear simulation using the complete DATCOM database. A comparison of performance is made with the traditionally designed JDAM autopilot, which makes extensive use of gain scheduling. To demonstrate the ability to tolerate a lack of wind tunnel data during control system synthesis, the same autopilot design is then evaluated in a nonlinear simulation that employs flight-validated, wind-tunnel-ba sed aerodynamic data. The autopilot is shown to successfully adapt by accounting for modeling errors and system nonlinearities throughout the flight envelope, eliminating the need for gain scheduling, and demonstrating the potential for dramatic reduction in dependence on wind tunnel data, which translates directly to a saving in development cost.