UE Interference Prediction from Dynamic Network Configuration
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Dynamic changes in network configuration, such as network node configuration and load, pose challenges for accurate interference prediction using machine learning in wireless communications systems.
Innovation Solution
A user equipment (UE) receives network configuration information via assistance signaling, which includes parameters like cell on/off status, duty cycle, and scheduling information, to enhance interference prediction using a machine learning model, and may switch between different models based on network changes.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If machine learning is used for interference prediction without network configuration information, then the prediction process is simpler, but the prediction accuracy deteriorates due to dynamic network changes
Solution Approach 1:
The network entity pre-configures multiple machine learning models at the UE, each trained for specific network configurations (e.g., different duty cycles, cell on/off patterns). This preliminary preparation allows the UE to quickly select an appropriate model without real-time training, balancing accuracy with computational efficiency in dynamic TDD systems
Solution Approach 2:
The system implements dynamic model selection where the UE monitors network configuration parameters (such as TDD uplink-downlink patterns, cell activity status) and switches between pre-configured ML models based on current conditions. This dynamic adaptation enables accurate interference prediction across varying network states without requiring a single complex universal model
2Measurement precision
If multiple machine learning models are maintained for different network configurations, then interference prediction accuracy improves, but device complexity increases
Solution Approach 1:
Instead of maintaining completely separate models for different configurations, the system uses parameter-based model selection where pre-configured models are associated with specific network parameters (duty cycle values, cell activity patterns). The UE selects models based on matching current parameter values, reducing complexity compared to managing fully independent models for each possible configuration scenario
Solution Approach 2:
The machine learning models are designed with universal structures that can handle multiple network configurations through parameter adjustments rather than requiring entirely different model architectures. This multi-functionality allows a smaller set of models to cover a broader range of network states, reducing the total number of models needed while maintaining prediction accuracy across diverse TDD patterns
Data Source
AI summary
Methods, systems, and devices for wireless communications are described. A device such as a user equipment (UE) may receive, via network signaling, an assistance information message that includes network configuration information corresponding to one or more inputs of an interference-prediction machine learning model. The network configuration information may include a set of parameters associated with one or more neighboring cell configurations. The UE may then generate, using the interference-prediction machine learning model, a channel interference prediction corresponding to one or more communication links at the UE based on the network configuration information included in the assistance information message. The UE may then communicate with a network entity based on the generated channel interference prediction.