Identification and control of nonlinear systems using multiple models based on the self-organizing map (SOM) [electronic resource] /
ABSTRACT: This thesis addresses the problem of modeling and controlling non-linear plants by utilizing a self-organizing map (SOM). It uses multiple models of the non-linear plant for identification, as well as a multi-model controller. The SOM is used to cluster the input space and it elegantly divides the space into regions that individually represents the local dynamics. The local models are then derived using least square fit for every SOM Processing Element (PE). Hence the global dynamics is represented by a set of local models. Multiple switching controllers are designed for these models using LMS algorithm. The switching between different controllers is performed by the SOM based on the present state of the system. The proposed methodology is tested on various nonlinear systems to demonstrate its performance. In the later part of the thesis, optimality is brought into controller design through adaptive critic methods. Dual Heuristic Dynamic Programming (DHP), a member of the adaptive critic family, is explored in detail and is implemented in the multiple model setting to design a globally optimal controller for nonlinear systems.
School:University of Florida
School Location:USA - Florida
Source Type:Master's Thesis
Keywords:adaptive control critic models multiple nonlinear som
Date of Publication: