Euclidean adaptive resonance theory with application to nonlinear and adaptive control systems

Pattern clustering is considered as one of the most challenging problems in the engineering field. Many researches were launched to tackle this problem. Fuzzy adaptive resonance theory (fuzzy ART) was one of the suggested solutions. It was presented as a clustering algorithm that employs some fuzzy operations in its learning rule. However, it shows an unacceptable behavior in clustering patterns of noisy background. It also exhibits the tendency to create clusters in an uncontrollable manner. Problems in classifying patterns are also considered as limiting factors to the said algorithm.;Euclidean adaptive resonance theory (EART) is the solution we suggest to the above-mentioned problems. In this research we build a clustering algorithm that has the ability to suppress noise levels that may exist in the pattern background. The suggested algorithm utilizes the Euclidean neighborhood criterion to decide the pattern belonging to a certain cluster. The learning rule of the suggested algorithm employs an averaging function to determine the location of the updated cluster in the clustering map. A similar Euclidean criterion is chosen to classify patterns, once training is accomplished. Clustering results showed an outstanding performance of the EART neural networks compared to their fuzzy ART counterparts.;EART is an unsupervised learning algorithm. It is analogous to the fuzzy ART one. We also introduce the EARTMAP algorithm. It is a supervised learning algorithm analogous to its fuzzy ARTMAP counterpart.;Both nonlinear and adaptive control fields are chosen as application platforms. Two target tracking experiments were conducted to show the applicability of the EARTMAP neural networks. The idea is to collect data that describe the behavior of the nonlinear and the adaptive controllers. The collected data are then used as EARTMAP training information. The said training yields EARTMAP neural controllers that substitute the original control systems.;EARTMAP controllers showed an excellent performance in doing the control job with high stability in behavior. The fuzzy ARTMAP, however, showed a less accepted performance with lower levels of stability.;ART neural networks are mainly applied in the fields of image and pattern recognition. This research introduces the EARTMAP and fuzzy ARTMAP as control systems.