The purpose of this project is to use Federated Learning techniques to train machine learning models for Aviation specific use case scenarios. Federated learning (FL) has shown potential for projects that involve data privacy and data isolation, while collaborating to achieve a shared objective. FL can create opportunities in the aviation industry for various stakeholders including, flight operators, weather services, emerging unmanned vehicle service providers, commercial space entities, or any other companies or institutions (with sensitive data) collaborating with NASA. In this initial project, we will explore technical feasibility leveraging available aviation data sets from commercial airlines, Federal Aviation Administration public data sources, autonomous aerial vehicles, and open-source organizations. The suggested dataset is DASHLINK (an open-source dataset provided by NASA) which is comprised of the last 160 seconds of a commercial flights. This is a timeseries dataset, containing 21 continuous features (airspeed, wind speed, AOA, pitch angle, flaps angle, etc.). The federated learning solution for this problem is a cross-silo, the silos are made from different airports (the airport is one of the features of the dataset), and the dataset is segmented as based on the name of those airports to mimic the real cross-silo setting.
· Pursuing a BS or MS in the following or related area: Computer Programming; Computer Science; Data Processing; Information Science/Studies; Aerospace / Aeronautical and Astronautical Engineering; Computational Science; Mathematics and Computer Science
· Strong AI/ML background
· Strong interest in Aerospace
· Preferably knowledge or experience working with federated learning models (such as TensorFlow Federated)