Deterministic Learning From Adaptive Neural Network Control for a Flexible Manipulator With Input Dead Zone

In this paper, a deterministic learning control strategy based on an adaptive neural network (NN) is proposed for a flexible single-link manipulator system with input dead zone. Firstly, a radial basis function NN is employed to estimate the system’s unknown dynamics, while an adaptive variable is used to mitigate the impact of input dead zone nonlinearity. Secondly, under the condition of partial persistent excitation, the system’s uncertain dynamics are accurately identified, and the learned knowledge is stored as constant NN weights. Subsequently, this knowledge is used to develop a learning-based controller that performs the same control tasks, thereby reducing computational burden and improving control performance. The Lyapunov direct method proves the system’s stability. Finally, numerical simulations confirm its accuracy and effectiveness.