Neural Network Closed-Loop Control Using Sliding Mode Feedback-Error-Learning
A novel variable-structure-systems-based approach to neuro-adaptive feedback control of systems with uncertain dynamics is proposed. An inner sliding motion is established in terms of the controller parameters. The outer sliding motion is set up on the system under control, the state tracking error vector being driven towards the origin of the phase space. The equivalence between the two sliding motions is shown. The convergence of the on-line learning algorithm is demonstrated and the conditions are given. Results from a simulated trajectory tracking control task for a CRS CataLyst-5 industrial robot manipulator are presented. The proposed scheme can be considered as a further development of the well-known feedback-error-learning method.
KeywordsHide Layer Sliding Mode Controller Parameter Variable Structure Control Lyapunov Function Candidate
Unable to display preview. Download preview PDF.
- 2.Shakev, N.G., Topalov, A.V., Kaynak, O.: Sliding Mode Algorithm for On-Line Learning in Analog Multilayer Feedforward Neural Networks. In: Kaynak, et al. (eds.) Artificial Neural Networks and Neural Information Processing. LNCS, pp. 1064–1072. Springer, Heidelberg (2003)Google Scholar