Keysight and Qualcomm Forge Digital Twin Path to AI-Driven 6G Networks

High-fidelity radio-frequency digital twins are emerging as a validation method for massive MIMO algorithms in 5G-Advanced and future 6G systems. Keysight Technologies and Qualcomm Technologies have demonstrated a workflow linking simulation, laboratory emulation, and over-the-air measurements to predict real-world network performance before deployment.

Why RF digital twins matter for 6G research
Massive multiple-input multiple-output (massive MIMO) is a core architecture of advanced wireless systems because it increases spectral efficiency by using large antenna arrays and dynamic beamforming. However, the performance of these systems depends strongly on site-specific radio-frequency (RF) propagation conditions. Algorithms that appear optimal in simulation may behave differently once deployed in complex urban environments or dense network topologies.

RF digital twins address this limitation by reproducing real radio environments in a physics-based simulation that integrates geospatial data, propagation models, and network parameters. Such models allow chipset designers, device manufacturers, and network equipment vendors to evaluate antenna configurations, beam management, and signal processing strategies before field deployment.

In the collaboration between Keysight and Qualcomm Technologies, the digital twin approach is applied to the development of massive MIMO systems intended for 5G-Advanced and early 6G research programs.

Linking simulation, lab testing, and real networks
The demonstration combines ray-traced RF propagation modeling with system-level testing and real network measurements. A photorealistic digital twin was created from Qualcomm Technologies’ massive MIMO prototype network deployed on its San Diego campus. The model incorporates site-specific radio conditions to reproduce realistic propagation scenarios.

The workflow integrates digital simulation, laboratory channel emulation, and operational network testing. Results generated in the simulated environment are compared with measurements collected from the prototype network and laboratory testbeds. The comparison evaluates several key performance indicators, including Reference Signal Received Power (RSRP), transmission rank, and data throughput.

Observed correlations between simulated and over-the-air results indicate that the digital twin environment can replicate real network behavior with sufficient accuracy for algorithm validation.

Generating datasets for AI-native radio access networks
A further objective of the project is the generation of high-quality RF channel datasets for machine-learning development. AI-driven radio access networks rely on large volumes of accurate channel data to train models used for network optimization and signal processing.

The validated digital twin workflow can generate datasets supporting techniques such as Channel State Information (CSI) compression, adaptive beam management, and AI-assisted precoding optimization. These capabilities are considered important for AI-native wireless architectures envisioned in future 6G systems.

Demonstration context
The joint workflow was demonstrated during Mobile World Congress 2026, held from 3–6 March 2026 at Fira de Barcelona Gran Via in Barcelona, Spain. The demonstration illustrated how digital twin modeling can connect algorithm design with measurable hardware performance across simulated and real network environments.

By aligning simulation, emulation, and field validation, RF digital twins provide a framework for evaluating massive MIMO architectures and AI-driven radio access network functions before large-scale deployment, reducing technical risk during the development of next-generation wireless infrastructure.

www.keysight.com