Impacto da Amostragem de Registos Experimentais de Relação Sinal-Ruído em Simulações ns-3

Abstract

In the last decade, we have seen the proliferation of Wi-Fi networks as a substitute for wired networks in different application domains, such as in the business and industrial environment, in the infrastructure of cities with a high population density, and in the home environment. The application scenarios are increasingly complex, such as the case of emerging networks that can be composed of wireless transceivers installed in Unmanned Aerial Vehicles (UAVs), Autonomous Underwater Vehicles (AUVs), balloons, etc. These very complex scenarios have driven the development of new network solutions. The idea of capturing real records of node positions and Signal-to-Noise Ratio (SNR), reproducing these conditions in Network Simulator 3 (ns-3), resulted in simulations with more realistic results. This approach was named Trace-based Simulation (TS), as it allows to reproduce past experiences even if the testbed is turned off, inaccessible, or no longer exists. However, we end up with thousands of SNR samples per second on each node involved in the experiment, which leads to the generation of records that take up several GigaBytes per experiment. Thus TS approach has some limitations. In this dissertation we study the impact of sampling SNR experimental records on the accuracy of ns-3 simulations based on the TS approach.