AI/ML Model Data Exchange via Dedicated Bearers for Operator Control
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing wireless communication systems lack a standardized mechanism for controlling and managing machine learning (ML) or artificial intelligence (AI) related data exchange, particularly for UE-side AI/ML model training, which raises concerns for telecom operators regarding data transfer security and resource usage.
Innovation Solution
Implementing a new ML-dedicated bearer with a standardized Quality of Service (QoS) Class Identifier (QCI) for AI/ML data exchange, enabling secure, controlled data collection, model transfer, and lifecycle management through a dedicated tunnel with configurable bit rates and QoS settings.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If a user plane-based data collection framework is used for AI/ML training data transfer, then data transfer flexibility and vendor implementation ease are improved, but network operator control and data security are worsened
Solution Approach 1:
The patent introduces a dedicated control plane as an intermediary between the user equipment and the data collection framework. This control plane establishes a trusted connection through standardized signaling procedures, allowing network operators to maintain control while enabling data transfer. The control plane acts as a mediator that authenticates and authorizes data collection requests before they reach the user plane.
Solution Approach 2:
The patent separates the control functions from the data transfer functions by implementing distinct control plane and user plane procedures. The control plane handles authentication, authorization, and configuration of data collection, while the user plane handles actual data transfer. This segmentation allows independent optimization of control and data transfer while maintaining operator control through the control plane.
2Productivity
If unknown data is transferred to unknown servers using existing infrastructure, then data collection capability is improved, but network security and operator control are worsened
Solution Approach 1:
The patent implements preliminary authentication and authorization actions through standardized control plane signaling before any data transfer occurs. The network operator pre-configures trusted server identifiers and authentication credentials, validating data collection requests in advance. This preliminary action ensures that only authorized data transfers to known, trusted servers are permitted, eliminating security risks from unknown servers.
Solution Approach 2:
The patent establishes feedback mechanisms through control plane signaling that provide continuous monitoring and control of data collection activities. The control plane receives status information about data collection and can dynamically adjust or terminate transfers based on security policies. This feedback loop ensures ongoing security verification and operator control throughout the data collection process.
3Device complexity
If standardized QoS control mechanisms are implemented for AI/ML data, then resource allocation control is improved, but system complexity is worsened
Solution Approach 1:
The patent leverages existing universal QoS mechanisms and signaling frameworks that are already standardized in wireless communication systems. By reusing established QoS parameters, message structures, and control procedures, the patent enables AI/ML specific resource control without creating entirely new complex systems. The existing QoS infrastructure is extended to support AI/ML workloads through configuration rather than structural redesign.
Data Source
AI summary
In accordance with example embodiments of the invention there is at least a method an apparatus to perform receiving or sending, between a network node of a communication network, and a user equipment information comprising a service request for a dedicated bearer for data exchange of at least one of machine learning or artificial intelligence related data, wherein the data exchange is for training at least one of a machine learning or an artificial intelligence model for a particular use case; communicating the information with the communication network; handling control over at least one of an artificial intelligence or machine learning related data exchange through a machine learning-dedicated bearer, wherein the artificial intelligence or machine learning related data exchange comprises: data collection, model transfer, and life cycle Management for at least one of a machine learning model or machine learning functionality control signalling.