Keysight Technologies and Qualcomm Technologies have launched a collaboration focused on the application of radio frequency (RF) digital twins.
The initiative aims to accelerate the development and validation of massive multiple-input multiple-output (MIMO) systems for 5G-Advanced and next-generation 6G wireless networks.
It is designed to allow industry stakeholders, including chipset, device, and network equipment manufacturers, to predict and optimise massive MIMO performance with greater accuracy before live deployment.
Both companies will present their workflow at Mobile World Congress 2026 (MWC 2026), detailing the construction of a photorealistic RF digital twin based on Qualcomm’s massive MIMO test network in San Diego.
The approach uses Keysight’s Channel Studio RaySim in conjunction with Qualcomm’s massive MIMO algorithms to model site-specific radio propagation effects. Validation of these models will be achieved by benchmarking results from the digital twin against measurements taken from Qualcomm’s end-to-end prototype network and laboratory-based testbeds using Keysight emulation solutions and Qualcomm devices.
This process aims to resolve discrepancies between simulated, lab-based, and real-world performance, particularly as site-specific RF conditions increasingly dictate deployment outcomes. Massive MIMO beamforming or precoding strategies frequently yield divergent results outside controlled environments.
Qualcomm senior director Kiran Mukkavilli said: “Realistic RF digital twins are essential for developing and validating advanced beamforming and AI-based optimisation techniques.
“By anchoring high-fidelity digital twins to Qualcomm Technologies’ end-to-end massive MIMO research platform, this collaboration demonstrates how high-fidelity modeling, combined with system-level testing and over-the-air validation, can improve confidence that massive MIMO algorithms will perform as expected in real deployments.”
The joint demonstration will evaluate correlations across technical metrics such as Reference Signal Received Power (RSRP), throughput, and rank, comparing them to over-the-air measurements for further validation.
The partners also indicate that this validated workflow is capable of producing high-quality channel datasets intended for training and validating AI/machine learning models. Targeted use cases include Channel State Information (CSI) compression, adaptive beam management, and AI-driven precoding optimisation, which are key enablers as AI becomes more embedded within radio access network (RAN) operations.
Keysight general manager and wireless solutions head Mosaab Abughalib said: “High-precision RF digital twins help engineers connect algorithm design to measurable hardware performance. By combining raytraced digital twins with advanced channel emulation, Keysight enables customers to predict massive MIMO behavior before deployment, reducing risk and accelerating innovation toward AI-native 6G networks.”
Keysight collaborates with Samsung and Nvidia on end-to-end AI-RAN validation
In a separate development, Keysight is working with Samsung Electronics and Nvidia on an integrated solution for end-to-end AI-RAN testing and validation. This initiative will also feature at MWC 2026.
The partnership leverages Keysight’s AI-RAN Simulation Toolset to automate the generation of realistic datasets, facilitate AI/ML model training, and unify benchmarking activities for AI-enhanced RAN modules. The initial focus is on PUSCH channel estimation.
The process addresses the common fragmentation in RAN validation workflows, where data science tasks are often siloed, which hampers reproducibility and comparative assessment prior to field trials. By coordinating dataset creation, model training, and benchmarking in a single pipeline, engineers can assess AI model behaviour consistently across diverse physical-layer scenarios that directly affect throughput and reliability.
The demonstration integrates the Keysight AI-RAN Simulation Toolset with Nvidia’s Aerial Testbed, utilising computing platforms such as NVIDIA GH200 and DGX Spark, and includes the NVIDIA Aerial Omniverse Digital Twin alongside commercial radio hardware from Analog Devices.
AI/ML models contributed by Samsung, Nvidia, and Keysight are said to be trained and evaluated within this unified environment to provide performance insights relevant to industry research and development teams.
Last month, Keysight introduced SOS Enterprise, an updated engineering data management platform designed to support AI adoption within semiconductor design at the enterprise level. The platform extends the capabilities of the existing SOS Core software by offering automated engineering data governance and traceability needed for large-scale AI deployment.
