Learning from Adaptive NN Dynamic Surface Control for Uncertain Pure-Feedback Systems

In this paper, deterministic learning (DL) from adaptive NN dynamic surface control(DSC) is investigated for pure-feedback systems with uncertain nonlinear dynamics. First, an adaptive neural DSC approach is utilized to ensure that tracking error converges arbitrary small around zero in a finite time. With the help of tracking convergence, the recurrent property of NN input signals (including the dynamic surface signals) is analyzed recursively, and a partial persistent excitation (PE) condition for the radial basis function neural networks (RBFNNs) is satisfied. Subsequently, accurate approximation of unknown system dynamics are achieved via DL. Then, a learning control(LC) scheme is proposed by reusing the learned knowledge to achieve superior control performance without any further readaptation online. Numerical simulations are given to illustrate the effectiveness of the proposed scheme.