Issues On Model-Free Adaptive Control Theory

This dissertation mainly focuses on some issues of the model-free adaptive control(MFAC) theory for a class of general nonlinear discrete-time SISO and MIMO systems.Regarding on the synthesis of several methods,such as the model-free adaptive control,sliding mode control,neural networks,the feedforward compensation and iterative feedback tuning(IFT),a series of new model-free adaptive control algorithms are proposed.These algorithms are designed to solve several problems such as,the tracking problem and disturbance rejection of nonlinear discrete-time SISO systems,the decoupling problem of the MIMO nonlinear discrete-time systems,and the cross-fusion research between MFAC and IFT.The main works and contributions are summarized as the following seven points.1.Based on the non-parametric dynamic linearization method for a class of general nonlinear discrete-time SISO systems,two similar non-parametric dynamic linearization methods are developed for a class of general nonlinear discrete-time MIMO systems.2.A series of new adaptive quasi-sliding mode control(ASMC) schemes are proposed for a class of nonlinear discrete-time systems.The tight format linearization (TFL) based linearization model of the system is adopted firstly.Secondly,based on this linearization model we propose the adaptive quasi-sliding mode control and give the theoretical analysis.Finally,another two similar adaptive quasi-sliding mode control schemes are put forward,which based on partial form linearization(PFL) and full form linearization(FFL),respectively.3.A class of neural network based adaptive quasi-sliding mode control algorithm (NN-SMC) is developed for the disturbances rejection and uncertainty problem to the general nonlinear discrete-time systems.The algorithm includes two parts:one is the design of TFL model based adaptive quasi-sliding mode controller,whose linearization parameters,i.e.pseudo-partial-derivatives(PPD) are estimated on-line from the I/O information of the system,the other is the estimation of the system uncertainty part acquired by a NN-based predictor.The BIBO stability is proven via rigorous theoretical analysis and the simulation results validate the effectiveness of the proposed method.4.A class of neural network based adaptive quasi-sliding mode decoupling control algorithm(NN-SMDC) is proposed for MIMO discrete-time nonlinear system.The algorithm also includes two parts:NN-based predictors and sliding mode controllers. The coupling effects among control loops are treated as the measurable disturbance which is estimated by NN-based predictors.Then the decoupling control of the MIMO systems is carried out with sliding mode control.The rigorous theoretical analysis is given also.5.In this part,an adaptive decoupling controller with feedforward compensation (AFCDC) is presented for a class of nonlinear multivariable discrete-time dynamic systems.This design is model-free,based directly on pseudo Jacobi matrix derived on-line from the input and output information of the system.Theoretical analysis and simulation results show that the model-free indirect adaptive decoupling control system is stable and convergent.6.Based on non-parametric dynamic linearization technique and IFT technique,a parameters' tuning method of the model-free adaptive controller(IFT-MFAC) is proposed for a class of nonlinear discrete-time systems.Three controller structures are deduced according to TFL,PFL,FFL theorem,then the controller parameters are tuned by IFT technique.The rationality of the 3 structures of the controller is assured by theorem of TFL,PFL and FFL,whereas,the controller structure is designed blindly by IFT.Obviously,the new method settles down this problem more properly.7.In this part,ASMC and NN-SMDC algorithm are introduced into the linear motor control system and the three-tank system,respectively.The simulation results show the validity of these control schemes.