Um método de filtragem de traços de mobilidade baseado em contato para simulações escaláveis de redes veiculares

Abstract

It is common to face scalability issues when running large-scale discrete-event simulations. This is a known challenge to everyone studying vehicular networks, where traces can have up to thousands of vehicles. As a solution, this work proposes a filtering technique for mobility traces that takes into account the contacts between vehicles. In opposition to common filtering techniques, the proposed filtering is done in a space-time manner in order to keep only vehicles of interest in the simulation, i.e., those participating either as source or destination of data transfers or those forwarding packets. A problem identified in traditional filtering that unintentionaly deletes traces is addressed by the proposed filtering method. A filter based on the proposed filtering is implemented in this work. The filter uses a modular structure in its implementation, which facilitates its future improvement and maintenance, and ensures its simplicity. The code for the tool used is made available and described throughout the text. Vehicular network simulations show that the results achieved with traces filtered using the proposed method converge much faster and exhibit reduced memory consumption compared to those obtained without the proposed filtering. Moreover, we show that it is possible to deliver a larger number of messages when the proposed technique is applied. This result indicates that the proposed filtering technique is better at maintaining the vehicle traces that participate in message routing when compared to the traditional technique.