Estratégia de otimização de roteamento baseada em jogo bayesiano para redes tolerantes a atrasos e desconexões com restrição de energia

• Ano de defesa: 2015
• Autor(a) principal: Maia, Sérgio Luiz de Freitas
• Tipo de documento: Tese
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Federal de Uberlândia; BR; Programa de Pós-graduação em Engenharia Elétrica; Engenharias; UFU
• Palavras-chave em Português: DTNs; Roteamento; Jogos bayesianos; Roteamento (Administração de redes de computadores); Algoritmos de computador; Routing; Bayesian games; CNPQ::ENGENHARIAS::ENGENHARIA ELETRICA
• Link de acesso: https://repositorio.ufu.br/handle/123456789/14357; https://doi.org/10.14393/ufu.te.2015.57

Currently, the research community in communication networks has given special attention to the study of emerging wireless networks such as sensor networks, mesh networks, ad hoc networks, pervasive computing systems and delay/ disruption tolerant networks (DTNs). The main feature of these networks is not require the presence of a communication infrastructure and, therefore, often present decentralized operation and auto configuration. Additionally, due to the DTN s highly distributed nature, it is desirable that it be also assigned to this type of network some level of awareness of energy consumption. Thus, in this Thesis, we propose an optimization strategy for a routing algorithm that was originally proposed without regard to the issue of energy constraint in delay/disruption tolerant networks (DTNs). The routing algorithm must use some utility function based on a number of different parameters (e.g., encounter history, mobility, sociability, etc.) to discover the better relay nodes. Our proposed strategy is based on modeling of the message forwarding as Bayesian game that aims specifically to capture the dynamic nature of the message replication decisions, the energy constraint of the devices and the uncertainty about the energy of other devices. An adaptive learning framework that allows the nodes to learn the optimal strategies over time is presented. We use a system for belief update about the energy of the other DTN nodes based on the accumulated observations of the destination nodes. Simulation results show that our proposed optimization strategy is able to lead the network to remain operational for a longer period of time and, consequently, to achieve a higher final delivery ratio.