AI-Driven AF Access to RAN Management Functions
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing communication networks lack efficient mechanisms for real-time, AI-driven influence on network functions, particularly in optimizing radio resource management and mobility management.
Innovation Solution
The implementation of an application function (AF) that utilizes AI and machine learning (AIML) models to influence network functions by interacting with network exposure functions (NEF) and network functions (NF), enabling direct access and parameter provisioning for real-time operations.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If AI/ML models are deployed in cloud-based network functions, then network resource optimization is improved, but network latency increases and real-time response capability deteriorates
Solution Approach 1:
The patent segments the AI/ML processing workload by deploying lightweight models at the network edge (in access nodes like gNBs) while maintaining heavier processing tasks in the cloud. This segmentation enables real-time local optimization decisions to be made close to the data source, reducing latency while still benefiting from cloud-based resource optimization capabilities.
Solution Approach 2:
The patent introduces a new spatial dimension to the network architecture by implementing a hierarchical structure with edge computing nodes positioned between user equipment and central cloud functions. This dimensional addition allows AI models to operate at multiple levels (edge and cloud simultaneously), enabling both real-time local response and comprehensive cloud-based optimization.
2Measurement precision
If centralized AI/ML processing is used for network function optimization, then model accuracy is improved, but system complexity increases and real-time capability deteriorates
Solution Approach 1:
The patent segments the AI processing functionally by implementing different model complexities at different locations: simplified models at the edge for real-time decisions and more complex models in the cloud for comprehensive optimization. This functional segmentation maintains overall system accuracy while reducing the complexity burden on any single component.
Solution Approach 2:
The patent introduces AI/ML training functions as intermediary components that operate independently at the edge and cloud levels. These intermediary training functions enable each node to maintain its own optimized models without requiring constant centralized control, thereby reducing system complexity while preserving model accuracy through localized adaptation.
3Adaptability or versatility
If network functions are exposed to application functions for direct access, then operational flexibility is improved, but security risks increase
Solution Approach 1:
The patent introduces network exposure functions (NEFs) as intermediary security layers between application functions and core network functions. These NEFs act as mediators that validate and authorize access requests, enabling operational flexibility for authorized applications while filtering out malicious requests and protecting the underlying network infrastructure from direct exposure to security risks.
Data Source
AI summary
The present application relates to devices and components, including apparatus, systems, and methods for an application function (AF) direct access to a radio access network (RAN) management function.