Research On Dual Adaptive Control

For the stochastic system with the uncertain parameter, designing adaptive controller is in the basis of Certainty Equivalence principle in many cases. So it is not the optimal control strategy using this method and can not eliminate the uncertainty of system parameter identification. Dual adaptive control can implement the optimal compromise with the good controlling and good estimating. Dual feature of controller express:i) tracking the desired output signal; ii)reducing the uncertainty of system parameters.The pole placement dual control of multivariable unknown parameter difference equation system is presented in this dissertation. Via Kalman filter, the unknown parameters are estimated; Based on CE principle, the pole placement is made to the previous system; The controlling signal which is obtained by the pole placement method is regarded as the desired input, the corresponding output is regarded as the desired output and the dual controller is designed with the bicriterion.Dual control of the stochastic system with the unknown parameter based on Maximum Mutual Information index is studied. The unknown parameters of the stochastic system are estimated via Kalman filter; Structure multi-objects optimal problem,then,adopt two-level optimal algorithm so as to obtain the approximate suboptimal solution of the dual control law.The variance minimization dual adaptive control approach with unknown parameters for the difference equation stochastic system is studied. The method which the difference equation transfer into state space model is presented, the unknown parameters can select from a finite model set. The optimal control law of every definite model can be obtained through Dynamic Programming principle, the suboptimal dual control law is obtained via weighting of posterior probability.The dual control of stochastic system with unknown parameter existing in the finite model set is studied. The multiple-step minimization variance performance index are replaced respectively by the one-step and two-step minimization variance performance indices, the one-step and two-step minimization variance control laws are resolved, the suboptimal dual control law is obtained by means of posterior probability weighting. The dual control of the uncertain model which is the stochastic system with polytope is studied. It is assumed that the polytope exists in the system matrix, input matrix and measure matrix of the state space model; The augmented system is obtained with posterior probability weighting; The suboptimal control law is achieved by the dynamic programming principle.The guaranteed cost control with output feedback controller for uncertain continuous-time and discrete-time stochastic system on norm-bounded is studied; The upper boundary general form of quadratic performance index of the closed-loop system is obtained; The sufficient and necessary condition of optimal guaranteed cost controller are addressed too make the closed-loop system stable asymptotically using Linear Matrix Inequality (LMI) technique.The non-fragile H∞controller via state feedback is studied for linear continuous-time and discrete-time systems with the norm-bounded parameter uncertainty in the basis of the T-S fuzzy model. The controller gains are assumed to have norm-bounded parameter uncertainties which can be either additive or multiplicative. The non-fragile H∞controller is presented via LMI.The visual toolbox is designed using GUI tool of Matlab, the simulating programs of dual control and guaranteed cost control studying in this dissertation are integrated into this toolbox.