Design And Research Of Adaptive Learning Control For Nonlinear Systems Subject To Constraint

Nonlinear systems have been widely used in production activities of human.Thus,mathematical model of nonlinear systems should be obtained in order to further carry out analysis on nonlinear systems.In the past decades,many modelling schemes have been proposed to obtain the accurate mathematical model,however,due to the structure of the system and the complicated motion,accurate mathematical models are exceedingly difficult to obtain or are too complicated to carry out further analysis,which cause the traditional control scheme to lead to poor control performances.In practice,actuator is often subject to nonlinear characteristics such as input deadzone and input saturation,which often pose tracking inaccuracies and even incur instability.Another problem to be considered is constraints on system states,i.e.,system states are subject to certain constraints,and if system states violate the prescribed constraints,operation insecurity may happen.Artificial neural networks,which are parameterized networks,have been widely used in image processing and model recognition due to the powerful ability of both approximation and generalization ability.Furthermore,Owing to simple structure,strong fault tolerance and adaptive learning ability,neural networks have also been widely used in control community.The paper solves the trajectory tracking problems for nonlinear systems subject to input constraints and state constraints.Neural networks are employed to approximate unknown parameters and unknown functions,then based on direct adaptive design method and backstepping design method,respectively,adaptive controllers are designed.A difference from traditional control schemes is that deadzone nonlinearity and asymmetrical saturation nonlinearity are directly approximated by neural networks and hyperbolic tangent function,respectively.Immesurable states and unknown disturbances are estimated by the designed state observer and the disturbance observer,respectively,and the controller is designed based on state feedback and output feedback.Constraint problems are solved by barrier Lyapunov function.The paper designs controllers for three kinds of nonlinear systems.Firstly,research on nonlinear theory is carried out for a class of nonaffine nonlinear systems with input deadzone,the sliding model surface is structured,and with state feedback and output feedback,respectively,the controller is designed,which makes all the system signals uniformly ultimately bounded and also illustrates it according to Lyapunov theory.Secondly,research on nonlinear theory is carried out for a class of multiple-input and multiple-output nonlinear systems with state constraints,and combing with barrier Lyapunov function,neural network approximation-based adaptive controllers are designed,which makes all the system signals uniformly ultimately bounded and also illustrates it according to Lyapunov theory.Finally,adaptive neural network controllers are designed for robotic systems,which are also considered as multiple-input and multiple-output nonlinear systems,based on state feedback and output feedback,respectively,eventually stabilizing the robotic systems and making all the system signals uniformly ultimately bounded.