Novel Interval Type-2 ANFIS Modeling Based on One-Step Type Reducer Algorithm

In this paper, a novel structure of Interval Type-2 Adaptive Network-Fuzzy Inference System (IT2-ANFIS) for modeling dynamic systems is proposed. Optimization algorithms are introduced to adjust the antecedent and consequent parameters of a fuzzy model. In order to avoid the classical iterative process commonly used in type reduction algorithms, a new one-step type reduction algorithm (OSTRA) is proposed, which in conjunction with IT2-ANFIS is tested on nonlinear dynamical system and one real system datasets. Furthermore, to validate the complete structure, experiments with numerical models are performed, in order to show the advantages of the proposed novel IT2-ANFIS structure, by obtaining better results than previously published T2-ANFIS structures. To illustrate a real application of the proposed modeling technique, a model obtained from a tractor steering wheel system was embedded in an electronic I/O board, to obtain a comparison of the on-line fuzzy model against the physical system, with satisfactory results in the approximation error.