A general purpose neuro-adaptive controller, which realizes an indirect-adaptive control strategy, is introduced. The proposed algorithm is based on the use of two Multi-Layer feed-forward Perceptron (MLP) Neural Networks (NNs), which are trained using a momentum back-propagation (MBP) algorithm. One of the MLP NNs is used to identify the process. The other MLP NN is used to generate the control signal based on the data provided by the NN identifier. Training is done on-line to tune the parameters of the neuro-identifier and neuro-controller that provides the control signal. Pre-learning is not required and the structure of the overall system is very simple and straightforward, no additional controller or adaptive signal is needed. Tracking performance is guaranteed via Lyapunov stability analysis, so that both tracking error and neural network weights remain bounded. An interesting fact about the proposed approach is that it does not require a NN being capable of globally reconstructing the nonlinear model. Several simulation examples are reported to demonstrate the merits of the proposed algorithm. As is shown in the simulations, the developed control algorithm can deal with different types of challenges that might happen in real-time applications, including the change of the reference model and the effect of applied unknown disturbances. The application of the proposed neuro-control algorithm to the adaptive control of electro-mechanical systems subject to stick-slip friction is shown in the last section of this paper. Reported simulations reveal that the proposed algorithm is able to eliminate the effect of this nonlinear phenomenon on the performance of the system. © 2010 Springer-Verlag Berlin Heidelberg.
CITATION STYLE
Mehrabian, A. R., & Menhaj, M. B. (2010). Stick-slip friction compensation using a general purpose neuro-adaptive controller with guaranteed stability. Studies in Computational Intelligence, 268, 179–203. https://doi.org/10.1007/978-3-642-10690-3_9
Mendeley helps you to discover research relevant for your work.