Previsão eficiente do posicionamento futuro de nós em redes móveis

• Autor(a) principal: Fynn, Enrique
• Data de Publicação: 2015
• Tipo de documento: Dissertação
• Idioma: por
• Título da fonte: Repositório Institucional da UFU
• Texto Completo: https://repositorio.ufu.br/handle/123456789/17825; https://doi.org/10.14393/ufu.di.2015.424
• topic: Ciência da computação; Redes remotas (Redes de computadores) - Previsão; Movimento - Previsão; Algoritmos de computador; MANET; VANET; Previsão de Posicionamento Futuro; Future Movement ForeCast; CNPQ::CIENCIAS EXATAS E DA TERRA::CIENCIA DA COMPUTACAO

The ability to predict where nodes might be in the near future may enable several new applications in a mobile ad hoc network (MANET). For example, content may be generated for an approaching potential consumer in pervasive computing scenarios or traffic jams may be predicted and prevented. We introduce PheroCast, a lightweight algorithm to do online predictions of a node’s future position based on its previous movement history. PheroCast, however, does not take into account the variations in the movement pattern along the day, using any previous history in the same way. For example, in a scenario where the same node travels every morning, but seldom in the evening, PheroCast would give the same or more weight to the data from the morning when predicting an evening trip, which would likely lead to a wrong prediction. Due to this limitation, we developed the Time of Day PheroCast, or ToD-PheroCast, an extended version of the original algorithm which takes the time of the day into account while making predictions, giving more emphasis to the history of movement within similar time windows. Finally, we evaluate the performance in three scenarios: (i) prediction of the position of buses in a metropolis, which are expected to have very regular mobility pattern; (ii) Taxis in a metropolis, which should lead to low accuracy predictions; and (iii) mobility of people interacting with wireless networks, that used traces collected by the author’s research group. Our evaluations show that ToD-PheroCast is up to 4.41% better than PheroCast in the bus scenario, in which it achieved over 85% accuracy in its predictions, and 0.72% better in the taxi scenario, in which the algorithm achieved up to 89.17% accuracy. Finally, in the wireless scenario, ToD-PheroCast achieved 81.02% accuracy. These results show that not only forecasting is possible in such scenarios, but that it may be done with high accuracy, online, and in a lightweight manner.