Georgia Institute of Technology
Employing Machine Learning Techniques to Increase the Quality of Ionospheric Modeling
Abstract
dc:description.abstractWireless communications are impacted by the ionosphere, which is the charged part of the upper atmosphere. This ionization affects the signal paths of communications, with some signals reflecting and others passing through. Understanding the state of the ionosphere, specifically the electron density, gives insight into how these signal paths are affected. Thus, knowing the electron density of the ionosphere is quite desirable. Because the ionosphere is difficult to densely measure, observations are sparse, and models are needed to complete the data. This work investigates modeling the ionosphere with Machine Learning (ML) and using various techniques to enable the usage of additional data types. This work begins by discussing the ionosphere and its impact on communications, followed by a discussion of machine learning models used to model the ionosphere, the techniques they employ, and the creation of a baseline model. Afterwards, I discuss how this baseline model can be improved by including historical information, including X-ray flux data that has not yet been used in state of the art 3D electron density ML models. I follow this with a discussion of how Total Electron Content (TEC) data, which is the integral of electron density, can be used as another input despite its irregular shape by using cross-attention, similar to transformers. This is expanded upon to outline how multiple different data types can be used in a single input sequence to allow for varied and unrestricted data inputs. Lastly, I introduce a method for training a 3D electron density model with TEC data alongside the normal training inputs to improve TEC approximation without compromising model outputs. The work is then concluded with suggestions for future direction and a brief discussion of the benefits of the various techniques.
Degree
thesis:*- Level thesis:degree_level
- Doctoral
- Department dc:contributor.department
- Electrical and Computer Engineering
- Grantor dc:publisher
- Georgia Institute of Technology
- Year dc:date.issued
- 2025
Author and committee
dc:creator, dc:contributor.*- Author dc:creator
-
- Smith, Liam Bruce
- Advisors dc:contributor.advisor
-
- Cohen, Alexander
- Anderson, David V.
- Committee members dc:contributor.committeemember
-
- Romberg, Justin
- Davenport, Mark
- Simon, Sven
Subjects
dc:subject × 5Rights
- Language dc:language.iso
- en_US
Identifiers
dc:identifier.*- Handle dc:identifier.uri
- https://hdl.handle.net/1853/78697