Multi-agent Planning in Dynamic Environments

/MASE-EGTI: Evolutionary Game Theory Agent-Based Simulator for Environmental Land Change

/Optimal Allocation of vBBUs Considering Processing Power and Distance Between MDC and RRH in Fog-RANs

/Multi-mode Intra Prediction for Learning-Based Image Compression

Palestrante: Leonardo Henrique Moreira

Orientador:Profa Célia Ghedini Ralha

Título: Multi-agent Planning in Dynamic Environments

Resumo: Dynamic environments can be affected by unexpected events that can lead to plan failures. Extensive research has been conducted to investigate the integration between planning and coordination in the presence of multiple agents, however, the third dimension of the problem related to execution issues have not been extensively addressed. The combination of these three dimensions under a generic approach is an open problem in the multi-agent planning (MAP) area. Related work propose recovery strategies that can be categorized into replanning and repairing, but they investigate research questions regarding the performance of strategies under restricted scenarios. In this research we propose a method for evaluating plan recovery strategies in dynamic environments considering the coordination complexity derived from the agents' coupling level. The method allows performance comparison of plan recovery strategies under different scenarios where planning can be carried out centralized or distributed with loosely or tightly coupled level. Plan recovery strategies will be applied in scenarios affected by weak to strong failures considering metrics such as plan length, planning time and simulation steps. Furthermore, we provide domain categorization and a MAP platform that supports the simulation and empirical analysis of different planning problems.

 

Horário: 14h30

Palestrante: Cássio Couto Coelho

Orientador:Profa  Célia Ghedini Ralha

Título: MASE-EGTI: Evolutionary Game Theory Agent-Based Simulator for Environmental Land Change

Resumo: In this work we present the MASE-EGTI to perform simulations of spatially explicit and agent-based models with interactional submodels based on evolutionary game theory (EGT). We depart from the Multi-Agent System for Environmental (MASE) simulation model presented by Ralha et al. (2013).  As well as MASE’s extention (MASE-BDI) presented in Coelho et al. (2016) – which employs the Belief-Desire-Intention mentalistic reasoning model as conceived by Bratman (1987).  Thus, MASE-EGTI puts forward an agent-based model that implements  a  wide  range  of  competing  assumptions consistent  with  the essence of EGT to allow analytical equilibrium ranges representing real-world scenarios.

 

Horário: 15h

Palestrante: Jonathan Mendes de Almeida

Orientador:Profa Célia Ghedini Ralha

Título:  Optimal Allocation of vBBUs Considering Processing Power and Distance Between MDC and RRH in Fog-RANs

Evento: IEEE Int. Conf. on Communications (ICC 2020)

 

Horário: 15h30

Palestrante: Henrique Costa Jung (mestrando)

Orientador:Prof.  Bruno Luiggi Macchiavello Espinoza

Título: Multi-mode Intra Prediction for Learning-Based Image Compression

Evento: IEEE Int. Conf. on Image Processing (ICIP 2020)

 

Profa Célia Ghedini Ralha (Este endereço de email está sendo protegido de spambots. Você precisa do JavaScript ativado para vê-lo.)

Coordenadora Seminários de Pós-Graduação em Informática 1-2020