NON-PDC Dynamic Output Feedback Control And Distributed H_? Filtering Of T-S Fuzzy System

As a universal approximator,T-S fuzzy model can approach nonlinear system with arbitrary precision in a compact set.According to the model's characteristics that the consequent parts of fuzzy rules are the linear dynamic systems,the mature and systematic linear sys-tem theory can be utilized to study the complex nonlinear systems.The parallel distributed compensation(PDC)method is to design the corresponding controller according to the IF-THEN fuzzy rules of the plant,so that the control premise part is consistent with the plant'fuzzy rules,and remarkable achievements have been made in the control synthesis problem of the system.However,there is an implicit assumption in PDC design method,that is,the premise variables of fuzzy rules can be measured.In the existing results,the system state variables or output variables often are selected as the premise variables,and the former are applied to a wider range of nonlinear systems.At this point,it is unreasonable to use the traditional PDC method to study the nonlinear systems with immeasurable states,and there are some limitations.Therefore,when the system state is immeasurable,the observer-based dynamic output feedback control of fractional-order T-S fuzzy system with single sensor and distributed H_?filtering problem in wireless sensor networks environment are studied respectively by using non-PDC method,making the design more reasonable,more wider application and higher flexibility.More,in this dissertation,the stochastic stability of frac-tional order systems with random jump factors is considered.Simultaneously,considering the distributed filtering problem under some factors such as data packet loss,saturation,noise interference,switching topology,Sigma-Delta quantizer,etc.that often occur in net-worked systems.Based on the above problems,the research results of this dissertation are as follows:1.For a class of fractional-order T-S fuzzy systems with order 0<?<1,a non-PDC dynamic output feedback controller based on observer is designed under the condition that the premise variables of IF-THEN fuzzy rules are immeasurable,and a non-quadratic fuzzy Lyapunov function is constructed to analyze the designed non-PDC dynamic output feed-back controller,where observer,controller and fuzzy Lyapunov function all depend on the estimated premise variables information.Under the assumption that the membership deriva-tives for a given compact set can be represented by a kind of weighted sum,the difficulty in knowing a priori the appropriate bounds of the membership derivatives to satisfy lin-ear matrix inequalities(LMIs)is removed.Using the matrix singular value decomposition method,new less-conservative strict LMI conditions are derived for the locally asymptotical stabilisation of the fractional-order T-S fuzzy system with immeasurable state variables.2.For a class of fractional-order nonlinear chaotic systems with random jumps,a fractional-order T-S fuzzy model with Markov jump is established.A non-fragile dynamic output feedback fuzzy controller is designed by non-PDC method under partially matched premises strategy.Moreover,the multiplicative random sensor noise over the measurement output is considered.Then,based on the matrix singular value decomposition method(SVD)and membership-function-shape-dependent(MFSD)analysis approach,new less-conservative sufficient conditions in term of LMIs are derived to guarantee the closed-loop fractional-order fuzzy system robustly stochastically asymptotically stable.3.The problem of non-PDC dynamic output feedback control is investigated for a class of fractional-order nonlinear systems with random jumps subject to any distribution under T-S fuzzy model with semi-Markovian jump,where the imperfectly matched premises(IMP)strategy is utilized to increase the rationality and flexibility of design.Simultaneously,actu-ator saturation in system operation is considered.A line-integral fuzzy Lyapunov function is first constructed for stochastic stability analysis of fractional-order system,avoiding the need for the upper bounds prior of the time-derivatives of membership function in general fuzzy Lyapunov function.In order to reduce the conservativeness of IMP method,relaxation matrices are introduced to obtain the sufficient conditions of the system stochastic stability which depend on the local information of membership functions in term of LMI.4.The distributed H_?filtering problem based on Sigma-Delta quantizer over wireless sen-sor networks(WSNs)is investigated for a class of discrete-time T-S fuzzy systems with immeasurable premise variables,where random data packet loss and multiplicative noise are considered.Using non-PDC method,a novel distributed filter dependent on the estimat-ed premise variables is designed,and a set of dual-subscript filter gains related to fuzzy rules are obtained.Simultaneously,treating the immeasurable premise variables as uncertainties,then a robust distributed filtering method is proposed for such an uncertain filtering error sys-tem.Compared with the conventional logarithmic quantizer,by utilizing the Sigma-Delta dynamic quantizer,the quantized measurement outputs are broadcasted to the distributed filters over WSNs,only requiring a finite number of quantization levels,and the static errors can be eliminated simultaneously.Based on fuzzy Lyapunov function,the less conservative mean-square stable conditions satisfying a prescribed H_?performance index are presented.Finally,a convex optimization problem satisfying a family of LMIs conditions is solved to obtain the filter gain parameters.5.The stochastic finite-time distributed H_?filtering problem is investigated for more gener-al Takagi-Sugeno fuzzy systems(TSFSs)with immeasurable premise variables over wireless sensor networks(WSNs)with switching topology.The practical factors including sensor saturation and measurement missing,which are modeled by mutually independent Bernoul-li processes with uncertain probability,are taken into account.Utilizing non-PDC scheme,a switching-type distributed filter based on estimated premise variables is designed to realize the sharing of filtering information and measurement information.A distributed robust fil-tering method is proposed to deal with uncertainties containing unknown premise variables.Then by constructing a model-dependent fuzzy Lyapunov function,new less conservative sufficient conditions in term of LMIs are obtained to ensure the distributed filtering error system stochastic finite-time bounded and achieving a modified H_?performance index un-der bounded disturbance.By solving a convex optimization problem satisfying a set of LMIs conditions,the filter gain parameters and average dwell time of topology switching signal are determined.