Closed-Loop Dynamics Identification and Robust Neural Control for Helicopter Systems via Deterministic Learning

In this article, a scheme combining closed-loop dynamics identification and output feedback neural control is investigated for a nonlinear helicopter system. First, in the learning phase using deterministic learning, an observer-based adaptive neural network (NN) control is designed. Under the premise of satisfying the persistent excitation conditions, the closed-loop dynamics model (knowledge) of the controlled system is accurately identified (learned), and the learned dynamics knowledge is stored in the NN. Second, to apply the acquired knowledge in constructing a robust controller with specified transient and steady-state performance, a novel learning-based sliding mode control (SMC) with prescribed performance is developed. Finally, based on a series of simulations and experiments, the results illustrate that the proposed learning-based SMC scheme exhibits excellent nonlinear approximation capability and robustness.