CENTRIC @ EuCNC & 6G Summit 2025 – Demonstrating Hardware-in-the-Loop Validation of AI-Based CSI Compression

The CENTRIC project made a high-impact appearance at the EuCNC & 6G Summit 2025, where it showcased a hardware-in-the-loop validation setup for AI-based Channel State Information (CSI) compression. This live demonstration underlined CENTRIC’s broader vision: to design and optimise the 6G air interface—including the physical layer and protocol stack—through automated, intelligent methods tailored to real-time use cases and user-specific conditions.

Vision: AI-Native Air Interface for 6G

CENTRIC aligns with the 6G vision where AI/ML-driven optimisation is foundational. While this vision is still evolving, early steps have already been taken in 3GPP Release 19 for 5G-Advanced, particularly in AI/ML-enhanced CSI feedback.

One promising use case is CSI compression using neural networks. In this approach, an autoencoder-based model compresses downlink CSI at the UE and transmits it via uplink to the gNB, which reconstructs it with minimal information loss—thereby reducing overhead and enhancing efficiency.

The Demo: Validating Keysight & Nokia’s AI-Based CSI Compression Model

This demo showcases a hardware-based validation setup for AI models performing CSI compression, developed in the CENTRIC project, including the following main elements:

• Multi-port RF transceiver (Keysight MTRX E6464A)
• RF Channel emulator (Keysight PROPSIM F8800A)
• PC with GPU for AI processing
• AI model under test (Nokia’s CSI Compression Model)

The CENTRIC project uses this setup to validate a low complexity transformer-based AI model for CSI compression developed by Nokia. The approach introduces a lightweight transformer-based neural network with an autoencoder architecture that achieves state-of-the-art performance with fewer parameters within the transformer-based architecture. As widely accepted in 3GPP, the system distributes processing by implementing the encoder on the User Equipment (UE) side and the decoder on the gNodeB side, enabling compression and accurate reconstruction of CSI across spatial and frequency dimensions. This architecture improves reconstruction accuracy while reducing CSI feedback overhead in uplink transmission. The model employs Normalised Mean Squared Error (NMSE) as a reconstruction metric for decompressed CSI.

Conclusion

The CENTRIC demo at EuCNC & 6G Summit 2025 powerfully demonstrated how AI-native techniques, tested with real hardware, are paving the way for efficient and adaptive 6G air interfaces. Through close collaboration with industry leaders like Nokia and Keysight Technologies, CENTRIC continues to validate innovations that will shape the networks of tomorrow.