Measuring, Understanding, and Estimating the Influence of the Environment on Low-Power Wireless Networks

摘要: After a decade and half of research in academia industry, wireless sensor networks (WSNs) are seen as key infrastructure able to monitor the environment which they immersed, thanks their miniaturization, autonomy, flexibility. Still, outdoor deployments WSNs (e.g., forests) notoriously difficult get right, partly due fact that low-power communication is greatly affected by characteristics target temperature, humidity, foliage). In absence quantitative evidence about application environments, asset drives successful reliable deployment experience gained from previous deployments, lab-like testbeds, or simulators however often do not resemble real-world environments. The general goal this dissertation support principled design improving understanding how natural affects network stack, providing tools modeling techniques address impact. This constitutes premise for be credible tool domain experts biologists) operating field. Our own practical need deploy WSN system wildlife monitoring mountains near Trento, Italy, pushed our goals towards oriented perspective, whose ultimate objectives are: supporting deployment; informing selection protocols, ensure well-suited environment; deriving models push envelope what can predicted simulated beforehand. To achieve these we must start first step—assessing quantitatively links in-field, i.e., where deployed. end, contribute with Trident Harpoon, in-field connectivity routing performance assessment principled, repeatable, automated, flexible collection measurements without tethered requiring coding end user. Then, using collect large set data traces six campaigns across different years, environments seasons, analysis quantified impact environmental factors on focusing primarily physical layers. Finally, exploit create both estimating link quality at run-time reproducing realistic conditions simulators. We argue expressly designed gathering empirical traces, characterization real modeling, together significantly advance state art rendering process designing deploying more repeatable predictable.