چكيده به لاتين
Today, with the progressive advances in processors, robotic science has also been changed so that the face of the world is changing. Vision is one of the most important senses in humans and animals. By using this feeling you can get a better understanding of the environment and conditions of it. In fact, using a camera in control systems means using a visual feedback error as feedback, and minimizing this error in order to use visual information in the control system. In this thesis, control of the nonlinear predictive vision model for controlling the moving robot is used. In order to optimize the cost function, recurrent neural networks have been used that can provide an answer without delay. Also, in the absence of a response in certain circumstances, it is capable of convergence into a subfield response. First, the discrete dynamical model of the mobile robot is considered as a prediction model, and the cost function coefficients at any given time are set so that the robot has the best performance along the path while maintaining constraint constraints. The need of the robot's precise parameters in the discrete dynamic model is considered to be a disadvantage of predictive controllers, which makes it impossible to accurately quantify these parameters in practice. In addition, robot parameters usually change over time. This defect in this thesis is enhanced by the method of robust model predictive control. Therefore, by using this method, it is no longer necessary to know the precise values of the robot system parameters, in addition to which the controller with a comparative property will be resistant to uncertainties such as parameter changes and environmental disturbances. In the following, controller stability was investigated using a predictive controller from the Lyapunov stability perspective. The simulation results show the efficiency of the proposed method for guiding nonholomonic moving robots. Finally, the sensitivity of the designed control algorithms to noise factors, parameter changes, and the proposed controller function are compared with a sample of recent methods presented in the papers.
