NWDAF Model Training for AI/ML-Based UE Positioning
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing wireless networks face challenges in enhancing data handling and network management to support accurate and reliable AI/ML-based positioning services, particularly in fifth-generation systems, with issues related to model training, data collection, and performance monitoring.
Innovation Solution
The implementation of a Network Data Analytics Function (NWDAF) with a Model Training Logic Function (MTLF) to train ML models for UE positioning analytics, enabling efficient data collection and model provision to the Location Management Function (LMF) for improved AI/ML-based positioning accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If AI/ML-based positioning models are trained using network data, then positioning accuracy is improved, but data collection and model training complexity increase
Solution Approach 1:
The system divides the AI/ML positioning system into distinct functional modules: NWDAF for data collection and analytics, MTLF for model training, and LMF for positioning execution. This segmentation allows each component to specialize in specific tasks, improving positioning accuracy while managing complexity through modular architecture.
Solution Approach 2:
The NWDAF acts as an intermediary between data sources and the MTLF model training function. It collects, processes, and prepares location measurement data from multiple network functions, then provides curated datasets to the MTLF for model training, reducing the complexity burden on individual components.
2Reliability
If real-time data collection is implemented for model training, then positioning reliability is improved, but network data handling load increases
Solution Approach 1:
The NWDAF performs preliminary data collection, filtering, and analytics processing before data reaches the MTLF for model training. By pre-processing location measurement data and preparing training datasets in advance, the system ensures reliable model training while reducing the real-time data handling burden on the positioning function.
Solution Approach 2:
The system implements continuous data collection and model training operations through the NWDAF and MTLF interaction. The NWDAF continuously gathers location measurement data from the network, and the MTLF continuously trains and updates positioning models, ensuring ongoing improvement of positioning reliability without overwhelming the network with intermittent bulk processing.
3Measurement precision
If specialized ML model training function is added to NWDAF, then positioning precision is improved, but system architecture complexity increases
Solution Approach 1:
The MTLF model training logic function is merged with the NWDAF data analytics function, creating an integrated system where data collection and model training occur within a unified architecture. This merging improves positioning precision by ensuring models are trained on high-quality network data while managing complexity through functional integration rather than separate distributed systems.
Solution Approach 2:
The NWDAF is designed with multi-functionality, serving both as a data collection analytics function and as a host for the MTLF model training function. This universal design allows the same network element to perform multiple critical functions for AI/ML-based positioning, improving precision while avoiding the need for additional specialized hardware or separate system components.
Data Source
AI summary
This disclosure describes systems, methods, and devices related to enhanced NWDAF. A device may perform inference for artificial intelligence (AI)/machine learning (ML) based positioning. The device may receive a trained ML model from a network data analytics function (NWDAF) containing a model training logic function (MTLF) for user equipment (UE) positioning analytics. The device may send a location measurement data request to an access and mobility management function (AMF) with an AI/ML positioning indication. The device may provide location measurement data or an analytics data repository function (ADRF) ID plus DataSetTag to the AMF.