Reinforcement Learning–Based Adaptive Attitude Control Method for a Class of Hypersonic Flight Vehicles Subject to Nonaffine Structure and Unmatched Disturbances

Abstract

This paper proposes a reinforcement learning–based adaptive attitude control (RLAC) method for a class of hypersonic flight vehicles (HFVs) output constrained nonaffine attitude control problems subject to unmatched disturbances. First, by considering the strong coupling of HFVs attitude dynamics, the uncertainty of aerodynamic parameters and the complexity of the flight environment, a second-order multivariable nonaffine nonlinear control system is obtained. Then, by introducing specific nonlinear function and coordinate transformation techniques, the output constrained nonaffine control problem is transformed into a stabilization problem of several new variables. Moreover, dual actor-critic networks and their adaptive weight update laws are designed to cope with unknown unmatched and matched structural uncertainties. Meanwhile, two super-twisting disturbance observers integrated with dual actor-critic networks are designed to compensate unknown unmatched and matched external disturbances. With the help of the Lyapunov direct method, output constraint, convergence of the estimated weights, and stability of the system are proved. Finally, the validity as well as superiority of the proposed method are verified by numerical simulations.