Navegando por Autor "Cavalcante, Tamer Stefani Guimarães"

Agora exibindo 1 - 1 de 1

Study about vehicles velocities using time causal Information Theory quantifiers.
(2019) Silva, Maurício José da; Cavalcante, Tamer Stefani Guimarães; Rosso, Osvaldo Anibal; Rodrigues, Joel José Puga Coelho; Oliveira, Ricardo Augusto Rabelo; Aquino, André Luiz Lins de
New proposals of applications and protocols for vehicular networks appear every day. It is crucial to evaluate, test and validate these proposals on a large scale before deploying them in the real world. Simulation is by far the preferred method by the researchers to evaluate their proposals in a scalable way with low costs. It is known, in vehicular network simulators, that realistic mobility models are the foremost requirement to make reliable evaluations. However, until then, the proposed mobility models are based on stochastic processes, introducing white noise in their formulations, which do not correspond to reality. This work presents the characterization of global, daily and hourly vehicles behavior through their velocities in different real scenarios. To perform this characterization was used the Bandt-Pompe methodology applied to time series from vehicular velocities. Then, the probability histogram was assigned to the following Information Theory quantifiers: Shannon Entropy, Statistical Complexity, and Fisher Information Measure. The application of this methodology, based on time causal Information Theory quantifiers, was possible to identify different regimes and behaviors. The results show that the vehicles velocities present correlated noise with f −k Power Spectrum ranging between 2.5 ≤ k ≤ 3 for highways traffic, 1.5 ≤ k ≤ 2 for mixed traffic, and 0.25 ≤ k ≤ 1 for denser traffic. Additionally, by using the same methodology, we verify that the mobility models used in simulation tools do not produce the same vehicular velocities dynamics observed in real scenarios, the best one presents a correlated noise with f −k Power Spectrum ranging between 0 ≤ k ≤ 2.5, for all traffic analyzed. These results suggest that these models must be improved.