Global dynamic neuroadaptive tracking control of strict-feedback systems

By mainly using the multi-switching technique, the author has proposed a neuroadaptive backstepping controller to ensure the global tracking stability of strict feedback systems recently. However, the problem of exponential growth of neuron numbers with the dimension of system states has not been solved yet. Regarding this, a dynamic neural network approach is proposed here to alleviate such a difficulty in this paper. The concept is simple in that only those neurons nearby the current state are activated for compensating the unknown nonlinearities and hence the number of activated neurons is significantly reduced. Instead of classifying the neurons into the passive and the active ones, a set of smooth activating functions are used to modulate the neural basis functions. The major benefit is that the resulting estimated functions are infinitely differentiable, and thus rendering the backstepping tool applicable. The proposed design preserves the global tracking stability and saves a lot of on-line computation time simultaneously. Simulation results demonstrate the validity of the proposed design.