A fuzzy logic-based approach is taken in this paper for modeling, prediction, and predictive control of unknown or uncertain chaotic systems. Only output data of the underlying system are required. A fuzzy predictive framework using a general structure of a linear combination of Gaussian basis functions is developed, where the basis functions are expressed as probability density functions and are empirically determined from the time-series data. A real-time one-pass learning algorithm is developed for identification of the chaotic system. Based on this framework, a fuzzy predictive controller is designed, which is especially suitable for sparse data in a real-time environment. Several simulation examples are then given for demonstration.
展开▼