• JoomlaWorks Simple Image Rotator
  • JoomlaWorks Simple Image Rotator
  • JoomlaWorks Simple Image Rotator
  • JoomlaWorks Simple Image Rotator
  • JoomlaWorks Simple Image Rotator
  • JoomlaWorks Simple Image Rotator
  • JoomlaWorks Simple Image Rotator
  • JoomlaWorks Simple Image Rotator
  • JoomlaWorks Simple Image Rotator
  • JoomlaWorks Simple Image Rotator
 
  Bookmark and Share
 
 
Master's Dissertation
DOI
https://doi.org/10.11606/D.3.2023.tde-15022024-105731
Document
Author
Full name
Rafael Molinari Cheang
E-mail
Institute/School/College
Knowledge Area
Date of Defense
Published
São Paulo, 2023
Supervisor
Committee
Sichman, Jaime Simão (President)
Sanchez, Maite Lopez-
Silva, Valdinei Freire da
Title in Portuguese
Um framework para concepção e aplicação de normas centralizadas para governar ambientes de aprendizado por reforço de incentivo misto.
Keywords in Portuguese
Agentes normativos
Aprendizado computacional
Frameworks
Jogos de incentivo misto
Abstract in Portuguese
Jogos de incentivos mistos compreendem um subconjunto de jogos em que os incentivos individuais e coletivos não estão totalmente alinhados. Esses jogos são relevantes porque ocorrem com frequência no mundo real, bem como em sistemas multiagentes, e seus resultados poderiam ser melhores para as partes envolvidas caso aspectos coletivos fossem considerados. Instituições e normas oferecem boas soluções para governar sistemas com incentivos mistos, mas na literatura, são usualmente estudadas e incorporadas de forma distribuída. Neste trabalho, propomos um framework para melhorar os resultados coletivos obtidos em ambientes de aprendizado por reforço multiagente de incentivos mistos. O framework propõe aprimorar o ambiente com um sistema normativo controlado por um agente externo de aprendizado por reforço. Ao empregá-lo, mostramos que é possível alcançar bem-estar social usando apenas arquiteturas tradicionais de agentes de aprendizado por reforço, mesmo em um sistema formado apenas por agentes egoístas.
Title in English
A centralized norm synthesis and norm enforcement framework for governing mixed-motive multiagent reinforcement lear.
Keywords in English
Mixed-motive games
Normative agents
Reinforcement learning
Abstract in English
Mixed-motive games comprise a subset of games in which individual and collective incentives are not entirely aligned. These games are relevant because they can be matched to frequently occurring events in the real-world, as well as in multiagent systems, and their outcomes could be better for the involved parties if collective aspects were considered. Institutions and norms offer good solutions for governing mixed-motive systems, but in the literature, they are usually studied and incorporated into the system in a distributed fashion. In this work, we propose a framework for reaching socially good outcomes in mixed-motive multiagent reinforcement learning environments by enhancing the environment with a normative system controlled by an external reinforcement learning agent. By employing this framework, we show that it is possible to reach social welfare using only traditional reinforcement learning agent architectures, even in a system of self-interested agents.
 
WARNING - Viewing this document is conditioned on your acceptance of the following terms of use:
This document is only for private use for research and teaching activities. Reproduction for commercial use is forbidden. This rights cover the whole data about this document as well as its contents. Any uses or copies of this document in whole or in part must include the author's name.
Publishing Date
2024-02-19
 
WARNING: Learn what derived works are clicking here.
All rights of the thesis/dissertation are from the authors
CeTI-SC/STI
Digital Library of Theses and Dissertations of USP. Copyright © 2001-2024. All rights reserved.