UC San Diego researchers launch free ‘digital twin’ for end-to-end testing of applications over wireless networks

Researchers at the University of California San Diego have developed an open-source “digital twin” of a wireless network: giving graduate students, startups and other innovators a free, easy-to-use way to test new technologies and get fast, realistic feedback. The platform could help accelerate the pace of wireless innovation.

“We are building a software replica of everything that happens when you use your phone, from the wireless signals traveling through the environment to the cellular network and apps that deliver data and services like video and Instagram,” said Dinesh Bharadia, associate professor in the Department of Electrical and Computer Engineering at the UC San Diego Jacobs School of Engineering, an affiliate of the UC San Diego Qualcomm Institute and senior author of the paper. “This will help industry and academia build new protocols and algorithms faster using software and AI, with less need for real-world experiments.”

Ali Mamaghani, a doctoral student in Bharadia’s lab and first author of the research, will present the platform, called “Tiny-Twin,” at the IEEE International Symposium on Dynamic Spectrum Access Networks on May 12.

Until now, researchers working outside large telecommunications companies have faced a difficult trade-off: realistic wireless testing has typically required either expensive hardware and proprietary data, or simplified simulations that fail to capture real-world conditions.

“Previously, when we developed new technology for 5G or 6G, the testing was done with a small-scale experiment or a small simulator, but the results were not very realistic,” said co-author Ish Kumar Jain, an assistant professor at Rensselaer Polytechnic Institute and alumnus of the UC San Diego Jacobs School of Engineering. “In addition, previous systems did not simulate the full protocol stack, the entire journey of data, from application to signal to delivery, not just the wireless signal.”

Finding the sweet spot

To bridge this gap, the team built Tiny-Twin on top of its existing open-source 5G research platform, called EdgeRIC, and rethought how wireless simulations are constructed. Instead of relying on a single model, the system allows researchers to plug in different types of wireless environments, including real-world measurements, standardized models and physics-based simulations.

The key advance is identifying how much detail is enough to be realistic while still running efficiently. Modeling every nuance of how signals bounce, reflect and distort in the real world can quickly overwhelm a computer. The team studied this trade-off and identified a “sweet spot” that preserves accuracy without excessive computation.

“It’s basically fidelity versus feasibility,” said UC San Diego electrical and computer engineering doctoral student and co-author Ushasi Ghosh. “We try to quantify that in this framework.”

The researchers also redesigned how the software runs, distributing computational tasks across multiple processor cores to improve speed and responsiveness.

“If you give one person too much work, it’s difficult to handle,” Mamaghani explained. “The same is true in computer systems. By distributing the computation across CPUs, we were able to reduce delays and operate closer to real time.”

Just as important, the platform was designed for accessibility. Tiny-Twin runs on standard computers using open-source software, with no need for specialized hardware such as GPUs or custom radio equipment. It also models the full network stack, allowing researchers to test complete applications, such as video streaming, rather than just isolated signal behavior.

A new tool for what’s next

The result is a flexible, realistic testing environment that opens new possibilities for wireless research and development.

“You can emulate a wide range of environments, from national parks to highways and complex urban settings,” said Ghosh. “It can support many different use cases.”

The platform may be especially valuable as artificial intelligence becomes more deeply integrated into research. AI models require large, labeled datasets and repeatable testing conditions, both of which can be difficult to obtain in the real world.

“This platform helps lay the groundwork for leveraging AI to develop wireless systems,” said Jain. “It provides a model where researchers can generate data and test algorithms under consistent, repeatable conditions.”

The team sees Tiny-Twin as a long-term endeavor rather than a one-off project. Future work includes using wireless signals to sense environments, detect interference or malicious transmitters and test AI-driven network control systems, such as smarter bandwidth allocation and improved video streaming performance.

In addition to Mamaghani, Ghosh, Jain and Bharadia, the paper, “Tiny-Twin: A CPU-Native Full-stack Digital Twin for NextG Cellular Networks,” authors of the paper included Srinivas Shakkottai of Texas A&M University. The Tiny-Twin code is publicly available at: https://github.com/ucsdwcsng/Tiny_Twin