الفهرس | Only 14 pages are availabe for public view |
Abstract During the recent decades, the state estimation of the dynamic systems and the state observation problem has been an active topic of research in different areas such as automatic control applications. Several conventional nonlinear observers have been suggested during the past decade and some of these are only applicable to systems with specific model structures. However, for most practical processes, defining an exact model is a hard task or is not possible at all. So, the main objective of this thesis is developing an intelligent nonlinear observer-based adaptive control to improve the performance of the unknown nonlinear systems. In this thesis, the proposed observer is developed using a new structure and the proposed adaptive dynamic programming (ADP) algorithm based observer is developed using new structure to get stable controllers for nonlinear systems. The observer-based adaptive control for controlling the systems without a priori knowledge of the system dynamics gives high performance for nonlinear and time varying parameters of the system. In this thesis, two proposed observer-based adaptive controller structures are introduced. The first proposed observer-based adaptive control is developed using diagonal recurrent neural network (DRNN). The updating weights for the proposed DRNN observer are developed based on the Lypaunov second method. ADP is designed based on a critic DRNN, which is constructed to approximate the optimal cost function, which can be applied to systems with higher degrees of nonlinearity and without a priori knowledge of the system dynamics. The main objective of the first proposed controller is to make the system states go to zero where the initial values are set as non-zero values. The second proposed observer-based adaptive control is performed using reinforcement learning (RL) algorithm which the critic and actor parts are implemented based on quantum diagonal recurrent neural network (QDRNN). The proposed adaptive tracking neural network control guarantee the faster convergence due to the developed updated algorithm for the controller parameters, which is derived based on the Lyapunov stability. The stability analysis for all the proposed controllers is studied. The main objective of the second proposed controller is to make the states of the controlled system track the desired states. The proposed adaptive systems have been simulated and compared with other existing controllers in the previous publications. Simulation results indicate that the response of the proposed controllers have good performance compared with other existing algorithms. The proposed controllers have been designed and implemented practically using a microcontroller for controlling the speed and position of a DC motor. Practical results show good and significant improvement in the performance of the proposed controllers to respond the system uncertainties and external disturbances. |