O-DU/O-RU AI/ML Collaboration for Adaptive UL MIMO Beamforming
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current O-RAN 7-2x-based UL mMIMO systems face performance degradation in high mobility and interference environments due to outdated SRS channel estimates and reduced signal dimension, lacking mechanisms for AI/ML model collaboration and sharing between O-DU and O-RU for optimized network performance.
Innovation Solution
Implement AI/ML-enabled collaboration mechanisms between O-DU and O-RU to exchange AI/ML capabilities, train models through reinforcement learning, and share them with other network entities, utilizing federated learning to optimize performance in terms of energy and fronthaul bandwidth.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Quantity of substance
If weight-based dynamic beamforming is used to reduce fronthaul bandwidth, then fronthaul bandwidth is reduced, but throughput performance degrades in high mobility and interference environments
Solution Approach 1:
The patent implements dynamic beamforming method selection where the O-DU and O-RU collaboratively choose between weight-based dynamic beamforming and DMRS-based beamforming based on real-time channel conditions. The system transitions between different beamforming approaches dynamically to adapt to changing mobility and interference conditions, resolving the contradiction between fronthaul bandwidth savings and throughput performance.
Solution Approach 2:
The patent changes the beamforming parameters and methods based on environmental conditions. By monitoring channel characteristics and comparing performance metrics, the system adjusts the beamforming approach (switching between WDBF and DMRS-BF methods) to optimize both fronthaul bandwidth utilization and throughput performance under different operating scenarios.
2Productivity
If AI/ML collaboration mechanisms are implemented between O-DU and O-RU, then network performance is optimized, but device complexity increases
Solution Approach 1:
The patent segments the AI/ML model into separate components distributed between the O-DU and O-RU. The O-DU handles higher-level decision-making and model training, while the O-RU performs local inference and execution. This segmentation reduces the complexity burden on any single component while enabling collaborative optimization of network performance.
Solution Approach 2:
The patent introduces an AI/ML collaboration framework that acts as an intermediary between the O-DU and O-RU. This framework enables coordinated operation through standardized interfaces and protocols, allowing the systems to share computational tasks and insights without creating direct complex interdependencies, thus optimizing performance while managing complexity.
3Adaptability or versatility
If multiple beamforming methods are supported for algorithm selection, then adaptability to different scenarios is improved, but device complexity increases
Solution Approach 1:
The patent implements a dynamic selection mechanism that activates only the necessary beamforming methods based on current channel conditions. Rather than always running all possible algorithms, the system dynamically enables or disables specific beamforming approaches (WDBF, DMRS-BF, etc.) depending on mobility and interference levels, thereby maintaining high adaptability while reducing active complexity.
Solution Approach 2:
The patent changes the operational parameters of beamforming methods based on environmental conditions. By monitoring channel characteristics and adjusting which beamforming algorithms are active and their specific parameters, the system achieves versatile adaptation to different scenarios while managing computational complexity through conditional execution.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
An optimized Open Radio Access Network (O-RAN) system implementing O-RAN split option 7-2x-based uplink (UL) multiple-input multiple-output (MIMO) operation includes: an O-RAN Radio Unit (O-RU); an O-RAN Distributed Unit (O-DU); and an artificial intelligence or machine learning (AI/ML) module comprising an AI/ML model and associated configurations in at least one of the O-RU and the O-DU. The at least one of the O-RU and the O-DU is configured to: i) determine a beamforming method and associated parameters for at least one endpoint associated with the O-RU; ii) receive measurement data from other one of the O-RU or the O-DU; and iii) modify at least one of the AI/ML model and the associated configurations based on the received measurement data.