Tsukada Lab. Computer Network Group
← Research v2xmachine learning

Tsukada Lab Cellular Communication Prototype

An End-to-End 5G System on Software-Defined Radio

We have released a five-minute narrated video of the cellular communication prototype now running in our lab. It shows a complete 5G system, built entirely on software-defined radio, from the hardware on the bench to two live over-the-air experiments.

A full 5G network on one lab bench. The prototype brings together the three parts of a cellular network: a 5G core network, a base station (gNB) and a user device (UE). All three run OpenAirInterface, an open-source software stack that follows the 3GPP 5G standard. The radios are two USRP X310 units, each with two UBX-160 daughterboards, linked by a 2×2 MIMO connection over the air. GPS-disciplined oscillators keep the timing precise on both ends.

Because every layer is software and open source, we can modify the 5G protocol stack anywhere, from the physical layer up to the core network. That is what makes the system a research platform rather than a fixed demonstration.

Two live demonstrations. In the first demo, we block the line of sight between the base station and the user with a board. The signal-to-noise ratio (SNR) on the monitor drops at once, and recovers as soon as the board is removed. The receiver log also shows the CQI, RI and PMI reports, the channel feedback that today’s 5G depends on.

In the second demo, an eight-wheel robot dog carries the user’s antennas across the floor. As it moves closer, the SNR climbs. This lets us emulate a moving user indoors, with repeatable trajectories.

Why we built it. Our main research line is AI-based, feedback-free MIMO transmission (Bo Qian et al., “CaFTRA: Frequency-Domain Correlation-Aware Feedback-Free MIMO Transmission and Resource Allocation for 6G and Beyond”, IEEE Transactions on Mobile Computing, vol. 25, no. 10, pp. 17653–17668, October 2026). A Transformer model learns a map from a user’s position to the radio channel, so the base station can choose its transmission settings without waiting for feedback.

The testbed lets us check those gains with real channels, measured over the air. Next, we plan to scale up the antenna array toward multi-user MIMO, take measurements outdoors with real vehicles, and eventually run feedback-free links for connected vehicles on our own test field.

We thank NI Japan for its hardware support and engineering guidance. This project is supported by JST ASPIRE.

Researchers and companies interested in collaborating on 6G testbeds are welcome to contact us.