Model-free Adaptive Control Based On Stochastic Approximation Algorithm

In order to solve the problems related to the dependence with the model of the system and the unmodeling system dynamics, model-free adaptive control method was proposed. Model-free adaptive control can implement the control only by the input data and output data without knowing the system model. Mode-free direct adaptive control is not model based theory; it is theoretically more close to the true meaning of control concept than other methods. Because of the avoidance of setting up the mathematical model, it is widely used in practice.Stochastic approximation method is suited to be used in model-free direct adaptive control due to its simple algorithm, free of the specific mathematic model of plant and the excellent ability to handle data with noise. Therefore, this paper pay much attention to the study of model-free direct adaptive control based on stochastic approximation algorithm. Through analyzing various control methods, the paper finds some exist problems and make some further improvements.In the process of studying the DA-NNC method, we find the bias of certain degrees exist in system tracking response. The proposed hybrid direct adaptive control with PID compensator can effectively reduced the existing error, and simulation result shows that this improvement is efficient. In addition, a further discussion on the importance of prefilter in the design is made while researching the virtual reference concept. We give the specific form of the filter, and proposed virtual reference model-free direct control method, and finally, the correctness and usefulness is proved through simulations.This paper also gives out some questions, which need further research and discussion, about the Model-free direct adaptive control based on stochastic algorithms.