Mavenir, Aira, and Intel Accelerate AI Benefits for Open vRAN at IMC24
The rising global demand for next-generation mobile services that optimizes communication networks and enables better user experiences has become increasingly critical. In parallel the emergence and integration of Artificial Intelligence (AI) and Machine Learning (ML) is enabling many industries to experience a revolution in improved performance, efficiency, analytics and general functionality.
Mavenir, Aira, and Intel are at the forefront of new and emerging AI/ML use-cases and have successfully demonstrated at India Mobile Congress 24 (IMC) as to how the integration of AI into Open Radio Access Network (Open RAN) technology addresses several major difficulties associated with the deployment and optimization of 5G networks. This partnership represents a key milestone in open vRAN innovation, bringing together Intel and Aira's cutting-edge AI and ML technologies with Mavenir's industry leading cloud-native Open RAN platform.
New Opportunities and Challenges of Enhancing Open RAN
The shift to 5G brings enormous opportunities, including increased capacity, better insights into user behaviour, and a path to 6G. However, Communications Service Providers (CoSPs) must deal with the associated difficulties of progressively disaggregated networks while optimizing Total Cost of Ownership (TCO).
As established leaders in the Open RAN space, Intel and Mavenir are committed to pioneering AI-driven solutions to fully realize Open RAN's potential, assisting operators in effectively navigating the associated challenges and providing new innovate use-cases.
Introducing Intel® vRAN AI Development Kit
The newly announced Intel® vRAN AI Development Kit, enables operators and developers to leverage AI within the RAN architecture improving network performance and functionality. The developer kit, which is built on the Intel® Xeon® platform, includes pre-trained AI models that have been tailored for RAN application scenarios. These models, created with Intel's experience and knowledge in AI/ML and telecommunications, provide a turnkey solution for incorporating advanced AI capabilities into RAN operations. Pre-validated AI inference code on Intel processors using Intel-optimized AI tools and frameworks like TensorFlow* Optimizations, OpenVINO™, and oneDNN allows operators to swiftly deploy AI upgrades to their networks.
The development kit was an essential component of the IMC demonstration, supplying critical architectural software elements necessary for the implementation of innovative use cases that address the growing need for enhanced performance, dependability, and quality of service and experience.
Figure 1. Intel vRAN AI Development Kit Framework
AI-enabled intelligent beam management Use-Case
Efficient beam management is critical in 5G, particularly at the cell edge, where channel conditions can seriously affect user experience, as seen in video streaming use-cases.
Mavenir integrates Intel's FlexRAN™ reference architecture and AI/ML models to alter beamforming weights and number of layers based on real-time channel information. The AI approach enhances throughput, decreases overhead and frees up CPU cycles by decreasing the frequency of channel status information (CSI) reports without sacrificing performance. This AI-assisted beam management ensures that consumers receive consistent high-quality service, even in tough situations while retaining exceptional quality of experience (QOE).
Improving Spectral Efficiency Using AI/ML-Based Channel Estimation Use-Case
Mavenir, in conjunction with Aira Technologies and Intel, have demonstrated the use of AI/ML technologies to increase spectral efficiency and lower the frequency of sounding reference signals (SRS) in time division duplex (TDD) systems. Aira's AI/ML model predicts the channel evolution between SRS slots, overcoming channel aging and resulting in higher SINR for UEs in MU-MIMO. Channel prediction also allows Mavenir's scheduler to make more accurate resource block allocation decisions. The overall result is a significant increase in throughput, particularly in high-mobility conditions, and improved overall spectrum efficiency.
Demonstration at IMC 2024
At IMC 2024, Mavenir, Aira, and Intel showcased how AI-powered technologies are advancing mobile network performance and reliability. The demonstration highlighted improvements in user throughput for massive MIMO (mMIMO) settings, highlighting the advantages of deploying Mavenir's virtualized Distributed Unit (vDU) software on Intel architecture. The demo also showcased how AI/ML models trained on 3GPP datasets maximize both beam management and spectrum efficiency in real time.
Figure 2. IML Based MIMO Channel Prediction Slide from IMC24 Demo
This technology collaboration and partnership, which incorporates capabilities into its cloud-native Open RAN solutions, is a crucial milestone in the growth of 5G and 6G networks, enabling new use-cases and higher levels of QOE.
Mavenir continues to expand its Open vRAN platform capabilities by leveraging the Intel® vRAN AI Development Kit and AI/ML models to achieve significant increases in network and user performance. This collaboration not only addresses current network issues, but also lays the groundwork for future AI-powered 5G and 6G mobile networks.