Sensoriamento participativo com descrição adaptativa das taxas de amostragem e consistência dos dados

• Ano de defesa: 2018
• Autor(a) principal: André, Carlos Henrique de Oliveira Monteiro
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Federal do Rio de Janeiro; Brasil; Instituto Alberto Luiz Coimbra de Pós-Graduação e Pesquisa de Engenharia; Programa de Pós-Graduação em Engenharia Elétrica; UFRJ
• Palavras-chave em Português: Engenharia elétrica; Redes veiculares; Sensoriamento participativo; CNPQ::ENGENHARIAS::ENGENHARIA ELETRICA
• Link de acesso: http://hdl.handle.net/11422/9480

Participatory Sensing (PS) is a paradigm of collaborative networks which provides incentives for users to participate of sensing tasks on a Region of Interest (RoI). In collaborative scenarios, one of the major challenges is to deal with mobile participating users who must balance the amount of data collected by each user so as not to impose an excessive load to the network and to them. The system must ensure that the data collected is reliable and that possible anomalous data does not influence the final result. In this direction, this work proposes a centralized system to adapt the sample rate assigned to each participating sensor, identifying the existence of inconsistency or unreliable data. It is assumed that part of the data can be collected by malicious or faulty sensors. Thus, it is necessary to evaluate the presence of inconsistent data, based on the mean and standard deviation of the samples. In the following, the sampling rate is calculated as a function of the variability of the samples collected in a given RoI taking into account the samples received and validated in the last time interval. The proposed system is evaluated using the dataset of the bus fleet of the city of Seattle, WA - USA, which records bus movements. At first, the robustness of the system is evaluated through simulations in the presence of inconsistent data. Results show that the system is robust to inconsistent data up to 70% of the nodes. In addition, results show the tradeoff between sampling rate and number of participating sensors. The more participating users, the lower the individual sample rate and the lower the amount of data transferred. It is possible to reduce approximately 67% of data load of the participants.