Support vector machines based neuro-fuzzy control of nonlinear systems

In this work, a novel neuro-fuzzy control structure has been proposed for unknown nonlinear plants, which is referred to as the SVM-based ANFIS controller since it has been emerged from the fusion of adaptive network fuzzy inference system (ANFIS) and support vector machines (SVMs). In the proposed controller, an obtained SVM model of the plant is used to extract the gradient information and to predict the future behavior of the plant dynamics, which are necessary to find the additive correction term and to update the ANFIS parameters. The motivation behind the use of SVMs for modeling the plant dynamics is the fact that the SVM algorithms possess higher generalization ability and guarantee the global minima. The simulation results have revealed that the SVM-based ANFIS controller exhibits considerably high performance by yielding very small transient- and steady-state tracking errors and that it can maintain its performance under noisy conditions.