Estimator-based dynamic learning from neural control of discrete-time strict-feedback systems

The dynamic learning issue from adaptive neural control for a class of discrete-time strict-feedback nonlinear systems is the main topic of this paper. Different from the traditional control schemes, a new auxiliary error estimator is constructed in this paper to promote the solution of weight convergence. Subsequently, a new weight updating law is designed based on the estimation error rather than the conventional tracking error. Based on the variable substitution framework, a new adaptive neural control strategy is constructed to assure the stability of the considered system, neural accurate approximation of unknown dynamics as well as the exponential convergence of neural weights. Such convergent weights are shown and stored as constants, i.e., experience knowledge. In light of the experience knowledge, a static learning control strategy is constructed. Such a control strategy avoids time consumption caused by updating weights, facilitates the transient control performance and lessens space complexity. Simulations are fulfilled to demonstrate the availability of the presented strategy.