Defesa Pública da Dissertação de Mestrado do Engenheiro Eletricista Kauã Minho Antunez
Data: 11/11/2021 – quinta-feira
Horário: 14h00min
A defesa será realizada através de videoconferênica, pelo link: https://mconf.ufrgs.br/webconf/00009898
Link alternativo para o caso de falha do primeiro:https://meet.google.com/zad-btmn-rzg
Banca Examinadora:
Profa. Dra. Lucíola Campestrini – PPGEE – UFRGS
Prof. Dr. Gustavo Henrique da Costa Oliveira – DELT (Deptº Eng. Elétrica) – UFPR
Prof. Dr. Diego Eckhard – PPGEE – UFRGS
Presidente da banca/Orientador: Prof. Dr. Alexandre Sanfelici Bazanella – PPGEE – UFRGS
Título da dissertação: “Design of Nonlinear controllers Through the virtual Reference Method and l1 Regularization”
Resumo: “This work proposes a new extension for the nonlinear formulation of the data-driven control method known as the Nonlinear Virtual Reference Feedback Tuning. When the process to be controlled contains a signifcant quantity of noise, the standard Nonlinear VRFT approach – that uses the Least Squares method – yield estimates with poor statistical properties. Thes properties may lead the control system to undesirable closed loop performances and even instability. With the intention to improve these statistical properties and controller sparsity and hence, the sistem’s closed loop performance, this work proposes the use of l1 retulatization on the nonlinear formulation of the VRFT method. Regularization is a component that has been extensively employed and researched in the Machine Learning and system Identification communities lately. Furthermore, this technique is appropriate to reduce the variance in the estimates. A detailed analysis of the noise effect on the estimate is made for the Nonlinear VRFT method. Finally, three different regularization methods, the third one proposed in this work, are compared to the standard Nonlinear VRFT.
Palavras-chave: Data-Driven Control, VRFT, Regularization”
