UE AI/ML Data Collection for MDT
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current wireless communication systems lack mechanisms to efficiently utilize Artificial Intelligence (AI)/Machine Learning (ML) capabilities at User Equipment (UE) for data collection and reporting, limiting the potential for improved network management and user experience.
Innovation Solution
The implementation of AI/ML data collection and reporting mechanisms within the UE, allowing the UE to perform local ML inference and report predictions to the network, thereby enhancing data collection efficiency and reducing the need for raw measurement reporting.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of information
If the UE reports raw measurement data to the network, then the network can perform comprehensive analysis, but the data transmission overhead and network processing burden increase significantly
Solution Approach 1:
The patent extracts only the essential inference results from the raw measurement data at the UE side, rather than transmitting all raw data. The UE performs local ML inference and reports only the processed results (e.g., predicted QoE metrics, channel state predictions) to the network, thereby extracting the valuable information while leaving the voluminous raw data at the source.
Solution Approach 2:
The patent introduces an intermediary ML model at the UE that acts as a mediator between raw measurements and network reporting. This intermediary processes the raw measurement data locally and transforms it into condensed inference results before transmission, reducing the data volume while preserving the essential information needed for network optimization.
2Measurement precision
If the network collects comprehensive measurement data from all UEs, then network optimization accuracy improves, but the signaling overhead and network processing complexity increase
Solution Approach 1:
The patent applies preliminary action by performing ML inference at the UE before data transmission. The UE pre-processes measurement data using locally stored ML models to generate inference results (such as predicted QoE, channel predictions) that are then reported to the network. This preliminary processing at the source reduces the complexity of network-side analysis while maintaining optimization accuracy.
Solution Approach 2:
The patent implements local quality by enabling each UE to perform specialized ML inference tasks locally based on its own measurement data and characteristics. Different UEs can run different ML models appropriate to their local conditions, generating locally-optimal inference results that are then aggregated by the network, rather than requiring the network to process all raw data uniformly.
3Productivity
If the UE performs local ML inference, then data transmission efficiency improves, but the UE processing requirements and energy consumption increase
Solution Approach 1:
The patent applies partial action by having the UE perform only specific, pre-configured ML inference tasks rather than comprehensive data processing. The network configures the UE with specific ML models and inference parameters tailored to the required measurements, so the UE performs only the necessary partial processing rather than exhaustive analysis, balancing energy consumption with productivity gains.
4Ease of manufacture
If the network configures MDT measurements at the UE, then measurement collection is standardized, but the flexibility to adapt to different AI/ML capabilities is limited
Solution Approach 1:
The patent introduces dynamics by making the MDT configuration adaptive rather than static. The network can dynamically configure different ML models, inference parameters, and reporting requirements based on the UE's reported AI/ML capabilities. The configuration can be updated over time as the network learns about UE capabilities and as ML models evolve, allowing the standardized MDT framework to adapt to diverse and changing AI/ML capabilities.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
A UE transmits a capability message to an access node in a wireless network, indicating capability corresponding to ML of the UE at a current time instance. The UE receives a configuration message for radio measurements, indicating set(s) of parameters to be measured. The UE receives a configured capability in a message and matches the configured capability with the capability of the user equipment at the current time instance. The UE measures the set(s) of parameters. The UE uses ML to infer value(s) for at least one of the parameters based on the measured set(s) of parameters and based on the received configured capability. The UE reports the inferred value(s) for the at least one parameter. An access node receives the capability message, determines and sends the configuration message and configured capability receives the reporting.