- Develops a new AI model to enhance RAN performance and throughput, increase system capacity, and reduce power consumption
- Collaboration enables Samsung to deploy AI-optimized RAN software globally
- Demonstration is facilitated by the AI-RAN Alliance and will be showcased at Mobile World Congress 2025
INDIA – Keysight Technologies, Inc. has collaborated with Samsung and NVIDIA to train artificial intelligence (AI) models for Samsung’s 5G-Advanced and 6G technologies. This enables Samsung to integrate a powerful AI model in their vRAN software solution, which will be demonstrated at Keysight’s booth, during Mobile World Congress 2025 (#5F41 Hall 5). The project is being developed at the AI-RAN Alliance as a work item.
Operators face significant challenges with traditional RAN performance, including limited throughput, high latency, and inefficient resource utilization. In commercial 5G deployments, uplink performance is often a concern, especially at the cell edge where limited user equipment (UE) transmission power affects the base station’s received signal-to-noise ratio (SNR). Conventional channel estimation algorithms struggle in low SNR regions due to excessive noise. AI modeling offers a new approach by enabling precise channel estimation, optimizing resource allocation, and reducing power consumption. This enhances system capacity, improves network efficiency, and delivers superior user experiences.
Keysight’s Channel Emulation Solutions provide channel generation capabilities for a wide variety of channel conditions, along with real-time signal processing and radio frequency (RF) capabilities. Samsung successfully generated advanced AI models for channel estimation in the uplink receiver, resulting in significant gains observed in the lab environment. For example, simulated experiments showed a 30% better cell edge throughput by using an AI model for channel estimation instead of the current static rules-based approach. The performance evaluation of this AI model was conducted using an end-to-end setup that included Samsung’s radio point and a distributed unit (DU) based on the NVIDIA AI Aerial platform. It was implemented on the NVIDIA GH200 Grace Hopper Superchip platform, and Keysight Channel Emulation and Core Emulation Solutions.
This accomplishment will drive innovation and pave the way for the widespread adoption of AI-enhanced RAN technologies.
Charlie Zhang, Senior Vice President at Samsung Research, said: “We are excited to realize these amazing performance benefits of AI-based channel estimation through close collaboration with Keysight, NVIDIA, and the AI-RAN Alliance. This is an important milestone for our industry, as we evolve toward a future of next-generation telecommunication networks that are AI-native and sustainable.”
Soma Velayutham, Vice President, Telecoms at NVIDIA, said: “With the power of AI, the radio signal processing pipeline can be much more flexible, unlocking significant performance and efficiency gains. Keysight’s AI-for-RAN innovation, in collaboration with Samsung and NVIDIA, is a significant step toward smarter, more reliable AI-native wireless networks.”
Giampaolo Tardioli, Vice President, 6G and Next Generation Technology at Keysight, said: “Our collaboration with Samsung, NVIDIA, and AI-RAN Alliance underscores the transformative potential of AI-native networks. This partnership not only enhances performance and reduces power consumption but also paves the way for more energy-efficient networks.”
