NWDAF AI/ML Training Trigger via MDT Data Collection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing 5G network data collection procedures are insufficient for training artificial intelligence/machine learning (AI/ML) models in gNBs or RANs, leading to inadequate model updating and performance.
Innovation Solution
Implementing a Network Data Analytics Function (NWDAF) that utilizes Minimization of Drive Tests (MDT) to collect RAN measurements, allowing for offline AI/ML model training or retraining based on collected datasets.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Quantity of substance
If existing 5G network data collection procedures are used, then data can be collected for model training, but the collected data is insufficient for effective AI/ML model training
Solution Approach 1:
The patent applies preliminary action by triggering MDT data collection before AI/ML model training is needed. The NWDAF determines when data collection is required and configures UEs to collect and report measurement data in advance, ensuring sufficient high-quality training data is available when model training needs to be performed.
Solution Approach 2:
The patent uses the NWDAF as an intermediary between the data collection process and the AI/ML model training. The NWDAF collects requirements from training entities, determines data collection needs, configures UEs via AMF, receives measurement reports, and provides the collected data to training entities, coordinating the entire data supply chain.
2Reliability
If comprehensive data collection is performed for AI/ML model training, then model performance improves, but signaling overhead and resource consumption increase
Solution Approach 1:
The patent applies local quality by configuring different UEs to collect different types of measurement data based on local conditions and training requirements. The NWDAF can specify particular measurement types, reporting formats, and collection parameters tailored to the specific AI/ML model training needs, rather than uniformly collecting all possible data from all UEs.
Solution Approach 2:
The patent implements partial action by collecting only the specific measurement data required for AI/ML model training through targeted MDT configurations, rather than collecting all possible network data. The NWDAF determines precise data collection requirements and configures UEs to collect only necessary measurement types, avoiding excessive data collection.
3Device complexity
If AI/ML model training is performed with insufficient data, then signaling overhead is reduced, but model updating capability deteriorates
Solution Approach 1:
The patent resolves this contradiction by performing preliminary data collection through MDT before model training is initiated. The NWDAF determines data collection requirements in advance, configures UEs to collect and store measurement data locally, and accumulates sufficient training data before the actual AI/ML model training occurs, eliminating the need for excessive signaling during the training phase.
Data Source
AI summary
Triggering of artificial intelligence/machine learning training in a network data analytics function is provided. A method for triggering artificial intelligence/machine learning training in a network data analytics function may include obtaining at least one machine learning model for training or retraining based on measurement data of a network. The method may also include determining that data collection is required prior to training or retraining the at least one machine learning model, and receiving one or more measurement reports that includes at least one dataset from the data collection. The method may further include determining whether additional assisted information from one or more network devices is required for training or retraining. The at least one machine learning model may be trained or retrained based on all collected datasets, which includes the at least one dataset from the data collection.