WindSDK: A Unified Framework for Querying Atmospheric Data
Keywords
Computational Fluid Dynamics, Atmospheric Boundary Layer, Turbulence, Weather, UAS, UAV, software
Presenter Abstract
Estimation of local atmospheric flow is important to inform glider pilots (or autonomous controllers) of ideal soaring conditions. There are two categories of methods that assist with this: 1). Using sensor data and real-time on-board computing algorithms to predict local windfields mid-flight, or 2). Using a combination of weather forecasts, CFD simulations, and machine learning to enhance off-board prediction capability, either preflight or mid-flight. In this work we introduce WindSDK, a tool capable of unifying outputs from all of the off-board compute capabilities into one user-friendly interface. This allows real-time querying of different predictive models to enhance glider path-planning and decision-making. Results show that WindSDK is useful in certain cases, like comparing windfields produced by different forecasts. Additionally, performance of WindSDK was analyzed as number of query points increased, and we showed that performance scales reasonably well, staying withing the same order of magnitude for query time while the order of magnitude of query points increased. In the final manuscript, we will lay out the general software architecture, describe some of the methods that allow flexible data-queries, and show some useful results from the software.
Presentations
Presented in Session 1: Platform Development I
WindSDK: A Unified Framework for Querying Atmospheric Data
Estimation of local atmospheric flow is important to inform glider pilots (or autonomous controllers) of ideal soaring conditions. There are two categories of methods that assist with this: 1). Using sensor data and real-time on-board computing algorithms to predict local windfields mid-flight, or 2). Using a combination of weather forecasts, CFD simulations, and machine learning to enhance off-board prediction capability, either preflight or mid-flight. In this work we introduce WindSDK, a tool capable of unifying outputs from all of the off-board compute capabilities into one user-friendly interface. This allows real-time querying of different predictive models to enhance glider path-planning and decision-making. Results show that WindSDK is useful in certain cases, like comparing windfields produced by different forecasts. Additionally, performance of WindSDK was analyzed as number of query points increased, and we showed that performance scales reasonably well, staying withing the same order of magnitude for query time while the order of magnitude of query points increased. In the final manuscript, we will lay out the general software architecture, describe some of the methods that allow flexible data-queries, and show some useful results from the software.
Presenter Biography (Optional)
Brendon Cavainolo is a postdoctoral research scholar at Embry-Riddle Aeronautical University at the CTO of Luminaflows (Advanced CFD Solutions, LLC.) . Brendon's research has spanned many subtopics in applied computational fluid dynamics, including multiphase flows (aircraft icing, CMAS/TBC interactions, water droplets), hypersonics (shock-drop itneractions), and atmospheric flows. Brendon received his Ph.D. from the University of Central Florida, and has worked at Battelle, the German Aerospace Center, and Lawrence Livermore National Laboratory.