Real Time Identification and Control of a DC Motor Using Recurrent Neural Networks
The paper proposes to use three Recurrent Trainable Neural Network RTNN models for real time DC motor system identification and state feedback-feedforward control. The proposed RTNN model have a Jordan canonical structure which permits to use the generated vector of states directly for DC motor feedback control. A Backpropagation through-time type learning algorithm for RTNN model training, is also described. The experimental results, confirms the applicability of the described identification and control methodology in practice. The given results of nonlinear mechanical system identification and control by means of three RTNN models show a good convergence and confirm RTNN qualities.
KeywordsRecurrent Neural Network Motion Control System Friction Compensation Real Time Identification Recurrent Neural Network Model
Unable to display preview. Download preview PDF.
- I. Baruch, T. Arsenov, S. Koynov: Recurrent Neural Network Models for Systems Identification and Control. Proc. of the 2-nd IFAC Workshop on New Trends in Design of Control Systems, Sept. 7–10, 360 (1997)Google Scholar
- Baruch I, Stoyanov I. and T. Arsenov: An Improved RTNN Model for Dynamic Systems Identification and Time-Series Prediction, Proc. of the 4th International Symposium on “Methods and Models in Automation and Robotics”, 26–29 August, Miedzyzdroje, Poland, Vol 2, 711 (1997)Google Scholar
- Y.H. Kim, and F. L. Lewis: High-level Feedback Control with Neural Networks, World Scientific Publ. Co. (1998)Google Scholar
- S. Weerasooriya and M.A. El-Sharkawi: Identification and control of a DC-motor using backpropagation neural networks. IEEE TEC 6, 663 (1991)Google Scholar