NSGA-II para a solução eficiente de problemas de roteamento multiobjetivo em redes de sensores sem fio

• Ano de defesa: 2019
• Autor(a) principal: Jeske, Marlon [UNIFESP]
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• dARK ID: ark:/48912/001300001m80k
• Idioma: por
• Instituição de defesa: Universidade Federal de São Paulo (UNIFESP)
• Palavras-chave em Português: Wireless Sensor Networks; Evolutionary Algorithm; Multiobjective Optimization; Redes De Sensores Sem Fio; Algoritmo Evolutivo; Otimização Multiobjetivo
• Link de acesso: https://sucupira.capes.gov.br/sucupira/public/consultas/coleta/trabalhoConclusao/viewTrabalhoConclusao.jsf?popup=true&id_trabalho=7697560; https://repositorio.unifesp.br/handle/11600/59853

In the last years it has been observed an increase in the use of applications and new technologies for wireless sensor networks (WSNs), which have the function of performing monitoring tasks in different environments. In this way, WSNs allow information from the physical environment to be connected to the internet, making them an essential part of the Internet of Things (IoT) concept. A WSN is composed basically of low-cost micro- devices with limited energy (namely sensors), able to collect, transmit and receive data. It is of fundamental importance to find ways and solutions that correspond to the intrin- sic needs related to the technological limitations of the network components, such as the maximization of the network energy efficiency (lifetime), minimization of the packet loss rate, maximization of connectivity coverage, among others characteristics called metrics of QoS (Quality of Service). This Master Thesis addresses a bi-objective routing problem in WSN recently proposed in the literature, which has as optimization criteria two con- flicting metrics of QoS, the most emphasized in the literature: residual energy efficiency and packet delivery reliability. The heuristic approach employed in the literature was not able to obtain results for large-scale environments. In addition, for small and large-scale environments, the literature heuristics provided slightly dense Pareto curves, making it difficult to evaluate the results. In this way, to solve the problem in this Thesis was deve- loped the multiobjective evolutionary algorithm Elitist Non-dominated Sorting Algorithm (NSGA-II). Simulation results show that the solution of the problem using NSGA-II has better efficiency in terms of solution quality and computational time when compared to literature heuristics and exact approaches. Besides that, due to the good performance obtained in small-scale environments through the evolutionary approach, it was possible to solve the bi-objective problem in large-scale environments with solutions in a viable computational time.