AvMet Applications, Inc. — National Aeronautics and Space Administration SBIR Phase I: A3

Description

<p style="margin-left:0in; margin-right:0in">We seek to reduce potential next-day weather impacts on the National Airspace System (NAS) through better delay prediction and better efficacy of Traffic Management Initiatives (TMIs) mitigating airspace congestion. Our innovative approach is to combine the retrospective method (use historical TMIs and delays on similar days for next-day prediction) with the predictive method (use fast-time simulation of next-day&rsquo;s weather, TMIs and resulting delays, based on the translated weather forecast) into a seamless ensemble.</p><p style="margin-left:0in; margin-right:0in">Until now, only the first part &ndash; historical analysis &ndash; was used for such predictions, and even then mostly short-range. However, AvMet has developed and validated a powerful weather-aware, superfast-time NAS simulator, DART, which can reproduce historically observed outcomes of a wide range of weather-impacted days with a good degree of accuracy and process a complex day of NAS operations in just over a minute. Our technology also has the potential to automate and accelerate similar-weather searches. This opens the door to an ensemble approach where both historical and simulation-driven forecasting methods are combined for next-day delay prediction and TMI recommendations.</p><p style="margin-left:0in; margin-right:0in">The methodology will be designed as a fully automated toolchain to work year-round, processing convective as well as non-convective (ceilings, winter weather, wind, etc.) weather impacts. It will use a number of innovative weather forecast translation, TMI simulation, and ensembling techniques. It will also use new and creative randomization methods to generate a range of potential scenarios based on next-day weather forecast.</p><p style="margin-left:0in; margin-right:0in">We believe that, by tuning the blend of retrospective and predictive ensemble methods, we should be able to utilize their strengths and mitigate their shortcomings, and demonstrate noticeably better predictive accuracy of this amalgamated ensemble vs. historical-only or simulated-only methods.&nbsp;</p>