With the potential to transform the future of global wireless networks, Rice University engineers are developing a cutting-edge testing framework to assess the stability, interoperability, energy efficiency and communication performance of software-based machine learning-enabled 5G radio access networks (RANs).
As 5G networks evolve toward more software-centric architectures, there is a critical need for advanced testing methods to ensure robust real-time performance. Funded by a $1.9 million grant from the U.S. Department of Commerce’s National Telecommunications and Information Administration (NTIA), the project aims to address this need by focusing on both communication and computing dimensions, considering the challenges posed by the inherent indeterministic behavior of such environments.
“Current testing methodologies for wireless products have predominantly focused on the communication dimension, evaluating aspects such as load testing and channel emulation,” said Rahman Doost-Mohammady, assistant research professor of electrical and computer engineering and the project’s principal investigator. “But with the escalating trend toward software-based wireless products, it’s imperative that we take a more holistic approach to testing.
“Our answer to this critical challenge is ETHOS, an innovative testing framework that not only evaluates communication performance but also considers the impact of computing environments and the intricacies of machine learning on RAN software.”
On Jan. 10, NTIA announced nearly $80 million in the third round of grants from the $1.5 billion Public Wireless Supply Chain Innovation Fund, which supports the development of open and interoperable wireless networks. Open and interoperable wireless equipment will help drive competition, strengthen global supply chain resilience and lower costs for consumers and network operators, according to federal officials.
“As part of President Biden’s Investing in America Agenda, the research and innovation supported by the Wireless Innovation Fund will bolster America’s global technology leadership,” U.S. Secretary of Commerce Gina Raimondo said. “The awards today will help stand up new facilities to usher in new wireless networks, ultimately leading to more jobs and lower costs for Americans.”
Following the creation of the new testing framework, the Rice researchers will conduct extensive testing on its efficacy and implement and deploy novel machine learning algorithms for 5G RAN on the NVIDIA-supported Aerial Research Cloud (ARC) platform — a fully programmable 5G network research sandbox designed to rapidly benchmark solutions through over-the-air networks.
In addition to working with NVIDIA, a global leader in visual and accelerated computing, the research team plans to engage existing industry contacts for feedback and fine-tuning of the framework, with a vision to expand collaborations as the project progresses.
“The broader impacts of this project are far-reaching, with the potential to revolutionize software-based and machine learning-enabled wireless product testing by making it more comprehensive and responsive to the complexities of real-world network environments,” said Ashutosh Sabharwal, the Ernest Dell Butcher Professor of Engineering, chair of the Department of Electrical and Computer Engineering and the co-principal investigator of the project. “By providing the industry with advanced tools to evaluate and ensure the stability, energy efficiency and throughput of their products, our research is poised to contribute to the successful deployment of 5G and beyond wireless networks.”
As key members of Rice Wireless, the researchers have extensive expertise in the latest wireless technologies. Doost-Mohammady is the technical lead for the Rice RENEW project, the world’s first fully programmable and open-source massive-MIMO platform. MIMO, or multiple-input multiple-output, is a wireless technology that uses multiple transmitters and receivers to transfer more data at the same time. Sabharwal is leading the Rice RENEW project and previously led the development of WARP, the world’s first software-defined MIMO research platform. Santiago Segarra, assistant professor of electrical and computer engineering and an expert in machine learning for wireless network design, is also a co-PI on the project.
