| We present our effort toward a better understanding of the novel aerial ad-hoc networking paradigm. In this paradigm, low-altitude unmanned aerial vehicles (UAVs) with wireless communication capability are used to assist networking among a set of ground stations. This new paradigm is attractive because UAVs can be dynamically deployed in a wide variety of geographical territories. Furthermore, UAVs often have high-quality, line-of-sight communication links with other UAVs and ground stations because aerial links usually suffer relatively little shadowing compared with their terrestrial counterparts. We also emphasize the use of commercial, off-the-shelf (COTS) components because of their high performance and economical cost, thanks to the positive cycle between market demand and technology advancement.; We further motivate this study by showing that the new paradigm can help solve performance problems due to interference in multi-radio, multi-hop networks. Although solutions such as antenna engineering and interference-robust medium access control can effectively mitigate multi-radio interference, multi-hop interference stands unsolved despite numerous attempts by the wireless research community. Using UAVs to ferry data can alleviate such interference by avoiding excessive multi-hop forwarding.; Field experimentation is indispensable to the understanding of the working of these new UAV networking systems comprising COTS components. For this reason, we have waged a measurement campaign using the Harvard UAV networking testbed. We have developed a set of efficient tools for trace collection and data visualization. We demonstrate the efficacy and advantages of our methodology with a number of sample experiment results and visualizations that reveal certain interesting interactions of the testbed with the environment that are otherwise hard to identify.; Last but not least, we seek to characterize the aerial link quality under various circumstances. Such link quality may fluctuate rapidly due to changes in UAVs' positions, velocities, as well as attitudes. Using regression and non-parametric statistical analysis, we compare and quantify the prediction capabilities of several existing models that take into accounts factors including distance, antenna gain pattern, antenna cross polarization, and, in some cases, ground reflection. We summarize the findings in a few compact guidelines for the designers of future aerial ad-hoc networking systems. We hope that our effort provides a concrete foundation upon which new airborne applications and networking protocols can be developed. |
