چكيده به لاتين
Prediction-based controllers have been widely applied to deterministic systems in recent years, however they have not been tried for stochastic systems with random noise.
In this thesis, the extension of the prediction-based (memory) controllers is considered for stochastic linear systems with input delay.
To do this, first the prediction of the state vector (prediction vector) is obtained from the system dynamics and the controller is constructed by using this prediction vector. This allows us to reduce (eliminate) the input time delay. Then by utilizing a suitable Lyapunov function, sufficient condition for obtaining the controller gain is obtained in the form of Linear Matrix Inequality (LMI).
In the following, the proposed approach is extended to the stochastic delayed systems with additive noise and external disturbance. To apply the proposed approach to the systems with additive noise, asymptotic stability is not suitable and hence, the notion of practical stability has been utilized. In addition, for attenuating external disturbance, robust H-infinity method has been used.
Since input and state delay may appear simultaneously in system dynamics, in the sequel the proposed method is extended to these kind of systems. With a similar approach, the prediction vector can be found for stochastic systems with state and input time delays and then it can be used to construct the controller. With this controller it is possible to reach asymptotic stability in the mean square.
The proposed approach has been applied to an active suspension system and a chemical process to show the effectiveness of the memory controllers in time-delay compensation of real-world systems.
