Research On Command Filter Based Adaptive Control Algorithm For Nonlinear Systems With Full State Constraints

The state,input,and output constraints of the system are all necessary constraints for ensuring stable operation of the industrial system.When these constraints are not met,the system may be affected,leading to performance degradation and even inability to operate normally.In addition to constraints,various uncertainties factors from measurement noise,model errors,model simplification,and disturbances also exist in nonlinear systems,which are also known as unmodeled dynamics and have a substantial influence on the stability of control systems.Control based on various unmodeled dynamics and constraints has become one of the hot and difficult issues in this field.Moreover,event-triggered control(ETC)technology has shown superior performance in energy saving and reducing transmission bandwidth utilization compared to traditional continuous control methods.In this thesis,several adaptive neural network tracking control methods are proposed using the command filtering control method for SISO and MIMO nonlinear time-varying full-state constrained systems,combined with unmodeled dynamics and ETC.The main contents of command filtering this thesis are described as follows:Firstly,an event-triggered control algorithm for a specific category of multi-input multi-output pure feedback systems,characterized by full-state time-varying constraints,combining the ideas of command filtering and adaptive neural networks.This algorithm uses a hyperbolic tangent function to map a constrained MIMO system to an unconstrained MIMO system.In order to solve complex computational problems,a dynamic surface control(DSC)strategy is utilized and command-filtered are used to replace the first-order filters in DSC.The command-filtered introduces an error compensation term to address the obstacles of the DSC method.This study focuses on a control scheme for the converted non-affine system,employing an adaptive event-triggered relative threshold strategy.The radial basis function neural networks(RBFNNs)are used to approximate unknown nonlinear function.Through Lyapunov stability analysis,the control strategy is validated,ensuring that all signals within the controlled system are semi-globally consistent and bounded.In addition,two simulations have demonstrated the feasibility of the scheme.Secondly,a finite-time unmodeled dynamic adaptive neural network command filtering control,is introduced for addressing the control scheme of non-strict feedback systems.Similarly,the error compensation term added to the command filtering is used to address the shortcomings of the DSC method.This study incorporates finite-time control,aiming to achieve system stability and control within a limited time.Dynamic signals are used to solve dynamic uncertainties.Standardized signal design is used to handle input unmodeled dynamics.Unknown nonlinear function are treated with RBFNNs.According to Lyapunov stability theory analysis and simulation,it is proven that the signal converges stably within a finite time,and the system is semi globally uniformly bounded in the closed-loop.