Banca de QUALIFICAÇÃO: LUIZ PAULO DE SOUZA MEDEIROS

Uma banca de QUALIFICAÇÃO de DOUTORADO foi cadastrada pelo programa.
STUDENT : LUIZ PAULO DE SOUZA MEDEIROS
DATE: 04/07/2024
TIME: 15:00
LOCAL: Sala Virtual
TITLE:

Multiobjective Optimization Model in Multiagent Systems with Reinforcement Learning

KEY WORDS:

Multi-agent systems, reinforcement learning, multi-objective optimization, decision process.

PAGES: 46
BIG AREA: Engenharias
AREA: Engenharia Elétrica
SUMMARY:

This work proposes a multi-agent multi-objective reinforcement learning algorithm for scenarios where multiple autonomous agents operate in a shared environment and must simultaneously optimize possibly conflicting objectives. The research aims to overcome the limitations of current approaches by providing adaptive, scalable and effective solutions to complex problems such as logistics planning and energy distribution. The objective is to explore approaches for the development of multi-agent reinforcement learning algorithms, proposing methodologies for learning in different interaction scenarios between agents, such as the use of multiple agents to accelerate learning, joint action learning, individual learning in environments with shared observation and individual learning in environments with individual observations. The relevance of the topic is due to the ability of reinforcement learning algorithms to adapt to changes in the environment, making them suitable for multi-objective optimization problems in real distributed decision situations. The methodology includes the development of a multi-agent reinforcement learning algorithm, the evaluation of the impact of different reward modeling techniques, such as Difference Reward and PBRS (Potential-Based Reward Shaping), and the comparison of the proposed algorithm with other solutions from the literature, considering aspects such as definition of a priori preferences between objectives, restrictions regarding the format of the Pareto Frontier, characteristics of communication between agents and the agents' level of knowledge about the environment. This research is expected to expand the state of the art in multi-agent reinforcement learning and multi-objective optimization, contributing to more effective and adaptive solutions to complex problems.

COMMITTEE MEMBERS:
Presidente - 347628 - ADRIAO DUARTE DORIA NETO
Externo à Instituição - FRANCISCO CHAGAS DE LIMA JUNIOR - UERN
Externo à Instituição - IGNACIO SANCHEZ GENDRIZ
Externo à Instituição - THIAGO HENRIQUE FREIRE DE OLIVEIRA - IFRN
Externo à Instituição - VINICIUS PONTE MACHADO - UFPI