BTT Missile Autopilot Design Using Neural Networks

The features of BTT control have determined that BTT missile is a strong coupling and time varying system, which make it difficult to design a control system. The traditional design method of three independent channels is not suitable. So developing an efficient method for BTT missile's control becomes a hotpot for researchers in the area of missile control. Two methods for autopilot design for bank-to-turn (BTT) missiles are presented, which meet the objective of the control design and prove to be efficient engineering method for the coupled nonlinear time-varying system synthesis.In first method, the architecture of adaptive critic autopilot is applied. Before I/O feedback linearization technique is applied to BTT missile, the nonminimum phase characteristic is circumvented by redefining outputs. The neural network consists of an associative search element (ASE) and an adaptive critic element (ACE). The ACE generate an internal reinforcement learning signal for tuning the ASE, which is employed to generate control action. An adaptive robust element is used to eliminate approximation errors and disturbances. The weight updating law derived from the Lyapunov stability theory is capable of guaranteeing both tracking performance and stability. Computing simulation results confirm the effectiveness of the proposed autopilot.The second autopilot is designed by combining the inverse system theory and adaptive critic neural network. In order to design the strong-coupled and time-varying control system for BTT missile, inverse system method is further studied. The nonlinear inverse transformation is realized by feedback linearization and inversion error is arised from imperfect modeling, approximate inversion or sudden changes in dynamics. Adaptive critic network is presented to attenuate the inversion error, the adaptation algorithm ensures that tracking error decays and the weights of the on-line neural network tend to constant values. Simulation results illustrate the performance of the autopilot.