Adaptive fuzzy modeling of hovering submarine based on on-line clustering

A fuzzy modeling approach based on on-line potential clustering is presented for a family of complex MIMO systems with severe nonlinearity and strong coupling such as a hovering submarine. The structure of the fuzzy model is defined as special group of If-Then rules with real constant consequents which is expressed as a feed-forward fuzzy neural network identified by on-line clustering and first order gradient algorithm. Using this method, the model structure and parameters can be achieved and updated rapidly and accurately, and the obtained rules of the model can be added, modified and deleted automatically according to the new information. In addition, the modeling approach using a structure with the fuzzy rules with constant consequents simplifies the modeling process compared to methods using the T-S model. Results of the simulation of hovering submarine demonstrate the effectiveness of this approach.