Banca de QUALIFICAÇÃO: RENAN JOSÉ DOS SANTOS VIANA

Uma banca de QUALIFICAÇÃO de DOUTORADO foi cadastrada pelo programa.
STUDENT : RENAN JOSÉ DOS SANTOS VIANA
DATE: 12/11/2020
TIME: 14:00
LOCAL: Videoconferência (Google Meet)
TITLE:

**

KEY WORDS:

Multiobjective Optimization, Dial-a-Ride Problem, Evolutionary Algorithms, Fit- ness Landscape Analysis.

PAGES: 105
BIG AREA: Outra
AREA: Multidisciplinar
SUMMARY:

Several problems encountered in science, engineering, industry, among other fields of knowledge, involve the simultaneous optimization of two or more often conflicting objectives. Problems with such characteristics are known in the literature as Multiobjective Optimization Problems. Due to the existence of a certain level of conflict between the optimized objectives, for this class of problems, there is no single optimal solution (global) as in mono-objective problems; instead, they produce a set of solutions that represent the possible trade-offs between objectives. Generally, in multiobjective optimization, the Pareto dominance relation is adopted to compare solutions. Thus, the optimal solutions to the problem are called Pareto-optimal solutions, and the image of this set in the objective space is called the Pareto-Optimal front. Over the past few years, both experimental and theoretical research has indicated that the increase in the number of optimized objectives promotes new and complex challenges. For example, the increase in the proportion of non-dominated solutions, the exponential growth in the number of solutions necessary to achieve a complete approximation of the Pareto-optimal front, the ineffectiveness of traditional solutions visualization techniques, difficulty in ensuring the maintenance of diversity population, among others. However, it is known that a multiobjective problem does not necessarily become more complicated when new objectives are added. Other properties of the problem can also increase its complexity, such as the characteristics of the test problems used. Also, combinatorial problems have not received as much attention as continuous problems, so it is still an area of research with questions to be explored. In this context, this work aims to promote a better understanding of combinatorial problems with many objectives and propose search strategies based  on each problem’s properties. For this purpose, two analyzes were performed using a Dial-a-Ride Problem (DARP) formulation with many objectives. Initially, an analysis of the fitness landscape was carried out to better understand the problem geometry and identify relationships between the objectives. Subsequently, a performance analysis involving some of the main multiobjective evolutionary algorithms was conducted. Both analyses showed that for the same formulation of the problem, different observations could be made both on the characteristics of the problem and the algorithms’ behavior when using different test sets. Finally, search strategies were proposed to deal with the increase in the number of objectives and other properties associated with the problem.

BANKING MEMBERS:
Interna - ELIZABETH FIALHO WANNER
Presidente - FLAVIO VINICIUS CRUZEIRO MARTINS
Externo à Instituição - LEONARDO CESAR TEONÁRIO BEZERRA - UFRN
Interno - MARCONE JAMILSON FREITAS SOUZA - UFOP
Interno - SERGIO RICARDO DE SOUZA