Este tópico é destinado a mostrar as pesquisa científicas em que já trabalhei. Serão citados trabalhos realizados, artigos publicados e trabalhos defendidos (graduação, mestrado, ...).
2011
Developing Strategies for Improving Planning and Scheduling of Actions in RTS Games
Autor(s): Augusto A. B. Branquinho, Carlos R. Lopes, Thiago F. Naves.
Resumo: This article describes techniques developed for generating and scheduling actions using partial order planning and SLA* in the production of resources for Real-time Strategy (RTS) Games. RTS games are characterized by two important steps. In the first step a plan of action should be carried out to produce resources. In the second step, the resources produced in the former step are employed in battles against the enemy. Resource production is vital to succeed in this sort of game. The developed algorithms significantly decrease the makespan, which is the time required for execution of the actions that achieve the goal for resource production.
Download: Link para download.
Publicação: 23rd IEEE International Conference on Tools with Artificial Intelligence, 2011, Boca Raton, Florida, USA.
Using Search and Learning for Production of Resources in RTS Games
Autor(s): Augusto A. B. Branquinho, Carlos R. Lopes, Thiago F. Naves.
Resumo: So-called real-time strategy games are characterized by two important steps. In the first step a plan of action should be carried out to produce resources. In the second step, the resources produced in the former step are employed in battles against the enemy. Resource production is vital to succeed in this sort of game. This paper describes an algorithm for planning and scheduling actions that provide the gathering of resources. Emphasis in this paper is given to the scheduling strategy, which is based on SLA*. In order to increase the efficiency of the scheduling, some strategies were developed to improve SLA*. Experiments show better results compared to other existing approaches.
Download: Link para download.
Publicação: WORLDCOMP - World Congress in Computer Science, Computer Engineering, and Applied Computing, 2011, Las Vegas - Nevada - USA. ICAI - International Conference on Artificial Intelligence, 2011.
2010
Planning for Resource Production in Real-Time Strategy Games Based on Partial Order Planning, Search and Learning
Autor(s): Augusto A. B. Branquinho e Carlos R. Lopes.
Resumo: Generally, a real-time strategy game is characterized by two stages. Initially, it is necessary to collect and produce resources. The next step is related to battles, taking into account the resources that were collected. The resources production stage is a key factor for winning the game. In this study the authors propose a mechanism for producing resources based on planning, supported by artificial intelligence using means-end analysis and scheduling. Emphasis is given to scheduling that uses an algorithm of real-time search and learning. The results show that the proposed system presents a better performance compared to related approaches.
Download: Link para download.
Publicação: Systems Man and Cybernetics (SMC), 2010 IEEE International Conference on, 2010, Istanbul - Turkey. IEEE Transactions on Systems, Man and Cybernetics., 2010. p. 4205-4211.
2009
[DISSERTAÇÃO DE MESTRADO] Planejamento de Recursos para Jogos de Estratégia em Tempo Real
Autor(s): Augusto A. B. Branquinho.
Resumo: Geralmente, um jogo de estratégia em tempo real é caracterizado por duas fases. Inicialmente, é necessária a coleta e a produção de recursos. O próximo passo está relacionado a batalha, levando em conta os recursos coletados. O estágio de produção de recursos é um fator chave para vencer o jogo.
Sendo assim, neste estudo é proposto um mecanismo de produção de recursos baseado em planejamento, suportado por inteligência artificial usando análises de meios-fim e escalonamento. O objetivo consiste em produzir planos para jogos de estratégia em tempo real. Ênfase é dada para o escalonamento usando um algoritmo de busca e aprendizado em tempo real.
São apresentados os resultados do sistema proposto comparando com trabalhos relacionados. Os resultados demonstraram que o sistema proposto apresentou uma melhor performance que as abordagens correlatas.
Download: Link para download.
Publicação: Dissertação de Mestrado. Universidade Federal de Uberlândia (UFU), Uberlândia, Minas Gerais. 2009.
Apresentação (Slides): Os slides da defesa do trabalho estão disponíveis em http://sdrv.ms/Tbv20C.
Uma Abordagem de Planejamento de Recursos para Jogos do Tipo RTS Baseada em Planejamento de Ordem Parcial
Autor(s): Augusto A. B. Branquinho e Carlos R. Lopes.
Resumo: In general, a real-time strategy game is characterized by two phases. First, it is necessary to produce (gather) resources. The next phase is related to battles taking into account the resources that were produced. The resource production phase is a key factor to succeed. In this paper the authors propose a mechanism for resource production based on artificial intelligence planning using means-end analysis and scheduling. A partial order planning algorithm was developed to satisfy the resource goals and priorities were assigned to its actions in order to achieve better schedulings. This is an important feature because a player must make a decision under time constraints. Results show that our system presents a better performance compared to related approaches.
Download: Link para download.
Publicação: Latin American Informatics Conference, 2009, Pelotas - RS. Latin American Informatics Conference, 2009.
Apresentação (Slides): Os slides da apresentação do trabalho estão disponíveis em http://sdrv.ms/QP7lNP.
2008
A Cryptographic Model Based on the Pre-Image Calculus of Cellular Automata
Autor(s): Gina M. B. Oliveira, Heverton B. Macêdo, Augusto A. B. Branquinho e Marcelo J. L. Lima.
Resumo: This paper is about the application of cellular automata (CA) in cryptography. In the approach investigated here, the ciphering is accomplished by pre-image calculus while deciphering is performed by CA temporal evolution. We investigated the application in a cryptographic model of the pre-image calculus algorithm proposed by Wuensche (Wuensche, A. 1992) known as reverse algorithm. The viability of this algorithm when applied to any arbitrary plaintext is based on the prerogative that all configurations of CA lattices have at least one pre-image. We speculate that transition rules with chaotic dynamical behavior are more probable to exhibit this characteristic. Therefore, we investigate if it is possible to find rule sets that guaranty the existence of one pre-image for all possible CA lattices. Theses rule sets were found by using a genetic algorithm (GA) which was guided by the forecast dynamical behavior parameter named as Z. The results of our experiments show that, beyond the dynamics forecast performed by Z, other two characteristics are important: the symmetric level (S) and the balance between two components of Z named Zleft and Zright.
Download: Link para download.
Publicação: AUTOMATA 2008 - EPSRC Workshop Cellular Automata Theory and Applications, 2008, Bristol. Automata-2008. Theory and Applications of Cellular Automata, 2008. v. 1. p. 139-153.
2006
[MONOGRAFIA DA GRADUAÇÃO] Software de Simulação Tridimensional de Mamas com Tumor
Autor(s): Augusto A. B. Branquinho.
Resumo: O estágio consiste no desenvolvimento de um Sistema de simulação tridimensional de mamas femininas a partir de imagens mamográficas. Possui como base a aprovação do Projeto de Pesquisa: “Localização Precoce de Câncer de Mama Utilizando Imagens Obtidas Através do Rastreamento Mamográfico”, pela Fundação de Amparo à Pesquisa do Estado de Minas Gerais.
Sob orientação do professor Bruno Nepomuceno Luiz o sistema foi desenvolvido utilizando conceitos de Processamento Digital de Imagens em Mamografias. O programa consiste em segmentar a região da mama e projetar sua imagem em um sistema tridimensional. Para o desenvolvimento do projeto foram utilizadas as ferramentas Microsoft Development Environment 2003 e Microsoft DirectX 9.0.
Download: Link para download.
Publicação: Defesa de graduação. Patos de Minas (MG). Centro Universitário de Patos de Minas (UNIPAM), 2006.
Apresentação (Slides): Os slides da apresentação do trabalho estão disponíveis em http://sdrv.ms/UXrhCF.
Segmentation of Breast Using Images of Digitized Mammographies
Autor(s): Guilherme B. Cunha, Adriano A. Perreira, Augusto A. B. Branquinho, Fábio J. Parreira.
Resumo: The breast cancer is one of more diagnosed between the women and one of the largest mortality causes by cancer in the feminine sex. The detection of the neoplasia can be made through of images analysis; this work considers the implementation of a sequence of stages for the mammas segmentation, starting from breast digitized images for the automatic location of neoplasias. The segmentation is one of the most important processes for the automatic location of breast cancer, because its accuracy indicates the success or the imperfection in this detection. The segmentation model proposed uses the binarization techniques, emphasis and segmentation of images. The binarization has been made using the algorithms of thresholding global and local adaptive. The emphasis of images was gotten using the erosion techniques, dilation and labeling. In the stage of the segmentation, it has been used the transformed of Watershed.
Download: Link para download.
Publicação: Proceedings, World Congress on Computer Science, Engineering and Technology Education, WCCSETE, 2006, Itanhaém. World Congress on Computer Science, Engineering and Technology Education, 2006. v. 1. p. 1237-1240.