UE Data Collection with RAN Awareness for AI-ML Training
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current data collection methods for AI-ML model training in wireless communication networks lack RAN awareness, making it impractical to perform UE-side model training due to limitations in data availability, storage capacity, computational capacity, and compilation capabilities.
Innovation Solution
The proposed solution involves a data collection framework with RAN awareness, where the UE receives data collection configuration from the RAN node, collects AI-ML model-related data, and delivers it through a data delivery tunnel to a UE server, while the RAN node accumulates and forwards the data to the UE server.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of information
If data collection is performed without RAN awareness, then the data collection process is simpler, but the data availability and quality for AI-ML model training are insufficient
Solution Approach 1:
The RAN node serves as an intermediary between the UE and the data collection framework. It receives measurement data from the UE, adds RAN-awareness information (such as radio conditions, network state), and forwards the enriched data to the training environment. This mediator approach enhances data quality without requiring complex changes at the UE side.
Solution Approach 2:
The data collection framework is segmented into multiple functional components: UE-side data collection, RAN-side processing and enrichment, and centralized training environment. This segmentation allows each component to focus on specific tasks, improving overall data quality while managing complexity through modular architecture.
2Manufacturing precision
If UE-side model training is performed, then model accuracy can be improved, but the UE lacks sufficient storage capacity, computational capacity, and compilation capabilities
Solution Approach 1:
The RAN node acts as an intermediary that collects and preprocesses data before transmission to the UE. By preparing data in advance and filtering out unnecessary information at the RAN side, the UE receives only the essential data needed for model training, reducing its computational burden while maintaining training effectiveness.
Solution Approach 2:
Data preprocessing, filtering, and organization are performed in advance at the RAN node before data reaches the UE. This preliminary action ensures that the UE receives ready-to-use, high-quality data, eliminating the need for complex data processing capabilities at the UE side while enabling effective model training.
3Reliability
If a structured data collection framework with RAN awareness is implemented, then data quality and relevance are improved, but the system complexity increases
Solution Approach 1:
The RAN node performs multiple functions within the data collection framework: it collects measurements from UEs, adds RAN-awareness information, filters and preprocesses data, and manages data transmission to the training environment. This multi-functionality approach consolidates complexity in the RAN node, which is already a sophisticated network element, rather than distributing it across multiple new components.
Data Source
AI summary
Apparatus and methods are provided for data collection with RAN awareness. In one novel aspect, the UE performs data collection with RAN awareness. In one embodiment, the UE receives from the RAN node data collection configuration, which configures AI-ML model related parameters for the UE, performs data collection and delivers the collected AI-ML model related data through the RAN node destined to a UE server. In one embodiment, the UE further receives a data collection request from the RAN node, a core network entity or the UE server. In another novel aspect, the RAN node performs data collection with RAN awareness. In one embodiment, the RAN node accumulates one or more sets of AI-ML model related data collected by one or more other UEs and delivers the one or more sets of AI-ML model related data together with AI-ML model related data collected by the UE.