Amazon Web Services, Virginia Tech Hume Center launch Emerging Technology Research Fellowship
The Cloud-based Distributed Radio Frequency Machine Learning project team members include (from left) graduate research assistant Zymmorrah Myers; Alyse Jones, co-principal investigator and research associate in the Virginia Tech National Security Institute Spectrum Dominance Division; William “Chris” Headley, co-principal investigator and associate director of the Virginia Tech National Security Institute Spectrum Dominance Division; and undergraduate researchers Joshua Lee, Akshay Pappu, Lynn Pham, Phillip Bozzay, Raj Kashikar, Anushka Trivedi and Edvin Gusinac. Photo by Isabella Rossi for Virginia Tech.
A student-led research team is working with Amazon to advance use cases for machine learning within the cloud for wireless communication applications.
The Virginia Tech National Security Institute is collaborating with Amazon Web Services (AWS) to give 11 undergraduate students and a graduate research assistant experience deploying state-of-the-art machine learning algorithms in the cloud for distributed radio frequency spectrum sensing through the Emerging Technology Research Fellowship. The fellowship aims to improve the performance of radio frequency spectrum sensing algorithms by leveraging multiple sensors collaborating through the cloud.
“We are thrilled to launch this fellowship with the Virginia Tech National Security Institute and work alongside students who will become tomorrow’s machine learning practitioners,” said David Appel, vice president of federal at AWS. “Spectrum sensing use cases have proliferated among our customers, and we look forward to demonstrating, through the work of these students, the value of cloud-based radio frequency machine learning to address them.”
Spectrum sensing is the process of
- Detecting the presence of transmitted wireless signals
- Estimating the parameters of the detected signals, such as duration, bandwidth, duty cycle, etc.
- Classifying the format of the detected signals, such as Bluetooth, Wi-Fi, cellular, etc.
- Fingerprinting the specific radio, or wireless devices, that transmitted the signal
Artificial intelligence and machine learning are of increasing interest within the domain of wireless radio frequency communications. Together, they will likely play an important role in the development of advanced technologies, such as 6G and NextG cellular systems, and the ability to counter them through electronic warfare.
Leveraging the National Security Institute's vast experience in research and development of radio frequency machine learning (RFML) solutions, the Cloud-based Distributed Radio Frequency Machine Learning project aims to develop novel approaches to pulling together data in the cloud from a variety of sensors. To streamline this process between the sensors and the cloud, the research team is developing an approach that will determine what information is sent to the cloud, how to wait for that information based on learned capabilities of the sensors, and what information should be provided back to the sensors to make better decisions in the future.
“Our team’s research is fundamentally about how to efficiently perform collaboration between the cloud and the RFML-enabled sensors and how to send the minimum amount of information to the cloud to reduce communication overhead,” said William “Chris” Headley, co-principal investigator and associate director of the institute's Spectrum Dominance Division. “It’s this well-known tradeoff between how much RFML data you send and the resulting performance, under the consideration of sensors with different capabilities, that is the crux of this work.”
The fellowship also will prepare undergraduate students for possible future careers using machine learning tools in support of national security use cases.
“The students are using industry-grade tools that they wouldn’t see or get to learn about in a normal classroom environment,” said Alyse Jones, co-principal investigator and a research associate in the Spectrum Dominance Division. “They have also been working very closely with the AWS team, so that gives them great insight into the industry.”
The fellowship expands on the Amazon-Virginia Tech Initiative for Efficient and Robust Machine Learning that began in 2022 under the direction of the Sanghani Center for Artificial Intelligence and Data Analytics. Tiasha Khan, program manager with the Hume Center, and Ehren Hill, the institute's associate director for education and outreach, are the fellowship’s co-principal investigator and principal investigator, respectively.
AWS team members have been involved in the project since the beginning, often attending meetings to offer feedback and advice, which students said has been impactful.
“They even offered office hours where we could ask more questions about the project or Amazon’s tools,” said River Thaboun, a first-year student majoring in computational and systems neuroscience. “That access to their team and their resources makes us feel like, even though we’re all undergraduate students, they want to put effort into us because they think the work we’re doing is important.”
The fellowship began at the beginning of the fall 2023 semester. Faculty and student researchers will continue their work this semester. Students will present the final project at the annual Hume Center National Security Colloquium in April.
The Ted and Karyn Hume Center for National Security and Technology is housed within the Virginia Tech National Security Institute and serves as the hub for national security-focused research, experiential student learning, and workforce development at the university. The summer internship program and similar efforts aim to work toward that effort of encouraging Virginia Tech students to consider a career in national security, the intelligence community, or related fields.
