A State Feedback Model-Free Predictive Controller for DC Motors Using Neural Network

In the automation control field, the model predictive controller is a modern controller with a simple control method, applied in the process of controlling the robot, motor,… In order to achieve desired quality, the control law is built based on the model datasets in the past, the present, and the future. The practical application of this method is hindered because it is so difficult to accurately determine model parameters. The paper proposes an advanced control method for position control of DC motors. The controller is combined from a neural network with an advanced model predictive controller instead of a classic predictive controller. In the first step, the technique employs a state-feedback control signal to stabilize the dynamical model of the system. Once the stable model has been obtained, in the second step, a proper neural network is designed with adaptive learning rules to learn the system behaviors. To realize the control objective, in the last step, a predictive control law is developed based on the online estimation results obtained from the network. The effectiveness of the proposed controller was carefully verified through in-depth simulation results.