Extension of the clustering identification by extending the Density Based Spatial Clustering of Applications with Noise approach to Multi-Input Multi-Output Piece Wise Affine systems: Application to an industrial robot

Abstract

In this paper the problem of clustering based identificationof a Multi-Input Multi-Output (MIMO) PieceWise Affinesystems (PWA) is considered. This approach, originallydesigned for systems with a Multiple-Input Single-Output(MISO) structure, is carried out by three main steps whichare data clustering, parameters vectors estimation and regionscomputing. Data clustering is the most important stepbecause the two other steps depend on the results given bythe used clustering algorithm. In case of MIMO PWA systems,we should cluster matrices of parametres which areconsidered high dimensionnal data. However, most of theconventional clustering algorithms do not work well in termsof effectiveness and efficiency since the similarity assessmentwhich is based on the distances between objects is fruitlessin high dimension space. Therefore, we propose an extensionof the DBSCAN (Density Based Spatial Clusteringof Applications with Noise) clustering approach for the identificationof MIMO PWA systems. The simulation resultspresented in this paper prouve the performance of the suggestedapproach. An application of the proposed approachto an industrial robot manipulator is also presented in orderto validate the simulation results. DOI: 10.61416/ceai.v25i2.8523