Defesa de Tese de Doutorado – Caio Merlini Giuliani – 14/8/2019
19/08/2019 14:44
| Defesa de Tese de Doutorado | |
| Aluno | Caio Merlini Giuliani |
| Orientador | Prof. Eduardo Camponogara, Dr. – DAS/UFSC |
| Data
|
14/8/2019 13h30 (quarta-feira)
Sala PPGEEL II (ao lado do Teixeirão) |
|
Banca |
Prof. Eduardo Camponogara, Dr. – DAS/UFSC (presidente);
Profa. Maria Aparecida Diniz Ehrhardt, Dra. – IMECC/UNICAMP; Prof. Juliano De Bem Francisco, Dr. – MTM/UFSC; Prof. Rodrigo Castelan Carlson, Dr. – DAS/UFSC; Prof. Eugênio de Bona Castelan Neto, Dr – DAS/UFSC (suplente). |
| Título | Contributions to Derivative-Free Optimization: an Exact Penalty Method and Decompositions for Distributed Control |
| Abstract: This thesis tackles the problem of applying decomposition methods in the area of derivative-free optimization of static and dynamical systems. A general purpose l1-penalty method is proposed for optimization in the absence of derivatives for objective and constraints. The method performs fast in numerical experiments. Often decomposition methods for optimization employ a multilevel optimization scheme. A new algorithm is proposed to specially handle the upper-level problems. The algorithm builds second-order models to better guide the search for an optimum. Regression is used to diminish the noise caused from the inexact solution of problems on the lower levels. Some formulations are proposed for the application of decomposition methods to the area of Model Predictive Control. | |
