Adaptive Backstepping-Based Neural Network Control for Hypersonic Reentry Vehicle With Input Constraints

In this paper, the attitude tracking control problem of hypersonic reentry vehicles is addressed by synthesizing a neural network (NN) using the backstepping control technique. The control-oriented model is formulated with mismatched and matched lumped uncertainties, which reflect the multiple aerodynamic uncertainties, external disturbances, and actuator saturation. Based on the universal approximation property of the radial basis function NN, an adaptive NN disturbance observer is developed to estimate the lumped disturbances online using only the tracking error state as its input vector. The "explosion of terms" problem in backstepping is avoided using a tracking differentiator. To address the input constraints, a sigmoid function is introduced to approximate the saturation and guarantee that the control input is bounded. In particular, a novel auxiliary system, driven by the tracking error and the input error between the unconstrained input and the constrained input, which was processed using the sigmoid function, is further designed to reduce the saturation effects and satisfy the stability requirement. Via modification of the adaptive laws, the tracking errors are guaranteed to be uniformly ultimately bounded based on Lyapunov theory. Moreover, several simulations are investigated to show the effectiveness of the proposed control scheme.