Digital Twin for 5G O-RAN
Digital twins are key enablers for AI-native radio access networks (AI-RANs), supporting data generation, policy validation, and what-if analysis without disrupting operational networks. However, their utility depends on how faithfully they reproduce cellular behavior, while wireless digital-twin fidelity remains poorly understood. We present the first systematic study of the sim-to-real gap in a cellular digital twin. We construct a twin of a live 5G O-RAN deployment by coupling a map-based ray-tracing channel simulator with a full-stack O-RAN platform and quantify channel- and network-level discrepancies using real-world measurements. We then develop Digital-Twin Parameter Optimization (DTPO), a Bayesian optimization framework that calibrates environmental and radio parameters. Experiments show that the calibrated twin cannot reproduce instantaneous channel realizations accurately but matches network-level behavior, with mean gaps of 2.3%, 5.1%, and 5.5% for per-user throughput, block error rate, and channel quality indicator.

5G Edge AI Offloading for VR Headset
VR headsets and AR glasses are fundamentally constrained by limited battery capacity and on-device computational resources. As a result, advanced AI applications on these devices increasingly rely on edge AI offloading. In this project, we developed a 5G O-RAN system with optimized user mobility management for delay-critical edge AI offloading using temporal graph embeddings. The accompanying video demonstrates a smooth human boxing application on a VR headset enabled by this 5G edge AI offloading pipeline: the headset streams camera images to a 5G edge AI server, and receives in real time the detected human bounding boxes along with the round-trip latency of the offloading process.

MSU's Private 5G Network
We built the first private 5G network on MSU campus. The network comprises six commercial indoor O-RUs, one commercial outdoor O-RU, and 20+ smartphones. The system is deployed on the third floor of MSU Engineering Building. This 5G network testbed supports both OpenAirInterface and srsRAN stacks. Four NVIDIA GPU A6000 devices have been installed for Near-RT RIC.

Measurement of Real-Time Uplink CSI in MSU's Private 5G Network
5G uplink Channel State Information (CSI) include not only MIMO channel matrix coefficients but also Time Advance (TA). Additionally, 5G operates in master-slave mode so the network can proactive hand over a smartphone from one gNB to another. This video shows the CSI measurement at a commercial base station from a smartphone when the user is standing and walking.

Facial Expression Reconstruction using 5G mmWave Signal
5G mmWave signals can be used to estimate human facial expressions. The left (blue) image shows the facial expression reconstructed using a depth camera, which serves as the ground-truth label. The right (orange) image shows the facial expression reconstructed solely from 5G mmWave signals using deep learning models.

Human Skeleton Reconstruction using Wi-Fi Signal
Radio signals from a commercial Wi-Fi router are used to estimate the skeleton of a walking person using an AI model. The overlaid green skeleton is generated by a camera using an off-the-shelf computer vision algorithm and serves as the ground-truth label for training the AI model. The overlaid red skeleton represents the skeleton estimated solely from the Wi-Fi signals.

3D Scene Reconstruction using 5G mmWave Signal
We built a bistatic ISAC device that uses 5G mmWave signals for sensing using RFSoC 4x2 and Sivers EVK. The sampling rate is 1.2288GSPS and the carrier frequency is 60GHz. The device operates in full-duplex mode: it transmits 5G signals for communication while simultaneously receiving the backscattered signals for sensing. As the device moves, it generates radio images of the surrounding scene in real time.

Bee Localization and Tracking using Tiny RF Backscatter Tag
A miniature backscatter tag, less than 4 mm in diameter, was designed and fabricated for bee tracking. The tag was attached to the thorax of a bee (see video), while an RF reader was installed outside the beehive. When excited by the RF reader, the tag generates a sub-harmonic response, which is used to localize and track the bee.