Native AI Network Architecture for 6G Edge Latency
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current 5G network architectures face challenges in supporting native artificial intelligence (AI) and machine learning (ML) due to latency issues, particularly with near-real-time and real-time requirements, and lack of efficient model management and data sharing across different vendors and network elements.
Innovation Solution
The introduction of a service-based architecture (SBA) for 6G networks that includes new network functions for unified management of AI/ML models, with data and model repositories located closer to the edge to reduce latency, enabling efficient sharing and collaboration between different vendors and network elements.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If data and model repositories are located in centralized core network, then network management is simplified, but latency increases for real-time AI/ML operations
Solution Approach 1:
The patent segments the centralized repository functions into distributed edge repositories and centralized management components. Edge repositories are deployed at network edges closer to AI/ML operations, while centralized management handles coordination and synchronization. This segmentation reduces latency for real-time operations while maintaining simplified centralized management for updates and coordination.
Solution Approach 2:
The patent introduces a hierarchical dimension to the repository architecture, organizing repositories across multiple levels (edge, regional, central). This multi-dimensional structure allows local edge repositories to serve immediate AI/ML needs with low latency, while centralized repositories provide overall management and coordination, effectively resolving the latency-management complexity trade-off.
2Adaptability or versatility
If standardized functions are implemented for legacy UEs, then compatibility is maintained, but support for native AI/ML operations is limited
Solution Approach 1:
The patent implements a universal network architecture that can perform both standardized legacy functions and native AI/ML operations through a common service-based framework. The same SBA infrastructure supports traditional network services while simultaneously enabling advanced AI/ML workloads, allowing the network to serve diverse requirements without sacrificing efficiency in either domain.
3Stability of the object's composition
If centralized model management is used, then model consistency is maintained, but real-time model updates and deployments are delayed
Solution Approach 1:
The patent implements preliminary action by pre-positioning AI/ML models in distributed edge repositories before they are needed for real-time operations. Models are prepared and staged at edge locations in advance, allowing immediate deployment when required without waiting for centralized retrieval, thus maintaining consistency while enabling rapid real-time updates.
Solution Approach 2:
The patent creates a dynamic model management system where models can be flexibly distributed to edge repositories based on real-time operational needs. The system dynamically adjusts model placement and updates across the distributed architecture, maintaining consistency through coordination while enabling rapid deployment responses to changing requirements.
Data Source
AI summary
The present application relates to devices and components including apparatus, systems, and methods to support native artificial intelligence model approaches in wireless communication systems.