Joint Spatial-Temporal Beam Management for High Mobility
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current beam alignment methods in new radio (NR) networks face challenges with high signaling overhead, particularly in high mobility scenarios, which compromises network performance due to the need for frequent beam measurement and reporting.
Innovation Solution
The implementation of joint spatial-temporal beam management using machine learning models, such as social LSTM, that predict the best beam or set of beams based on user equipment mobility patterns and interactions, reducing the need for full beam sweeping and thereby decreasing signaling overhead.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If frequent beam measurement and measurement reporting are implemented to guarantee continuous connection in high mobility scenarios, then connection reliability is improved, but signaling overhead increases
Solution Approach 1:
The system performs preliminary beam measurement and characterization during an observation period before the prediction period. Historical beam data is collected and used to train machine learning models that predict future optimal beams, eliminating the need for continuous frequent measurements while maintaining connection reliability through accurate predictions.
Solution Approach 2:
The patent replaces the mechanical process of continuous beam sweeping and measurement reporting with a machine learning-based prediction system. The ML model processes historical beam data and mobility patterns to predict future optimal beams, substituting the traditional measurement-based approach with an intelligence-based approach that significantly reduces signaling overhead.
2Measurement precision
If full beam sweeping is performed to ensure accurate beam alignment, then beam alignment accuracy is improved, but signaling overhead and measurement procedure overhead increase
Solution Approach 1:
Instead of performing full beam sweeping, the system conducts a limited observation period with sparse beam measurements. The machine learning model then infers future optimal beams based on this partial data and mobility patterns, achieving accurate beam alignment without the overhead of comprehensive beam sweeping.
Solution Approach 2:
The system creates a virtual copy of the beam management process through machine learning prediction. Rather than physically measuring all possible beams, the ML model copies the essence of beam selection by learning from historical data and predicting future optimal beams, reducing measurement overhead while maintaining accuracy.
3Loss of information
If machine learning-based prediction is used to reduce signaling overhead, then signaling overhead is reduced, but system complexity increases
Solution Approach 1:
The machine learning model serves multiple functions simultaneously: it predicts future optimal beams, characterizes user mobility patterns, and determines observation period duration. This multi-functionality consolidates what would otherwise require separate systems into a single unified model, managing complexity while reducing overhead.
Data Source
AI summary
The technology described herein is directed towards a joint spatial/temporal domain beam management including spatial domain beam management on the observation window and temporal beam management on the prediction window. Incorporating the spatial and temporal dependencies increases the prediction interval and makes the predictions more accurate because the spatial beam management is performed on the observation interval. Also described is interaction-aware multi-user equipment (UE) beam management technology that models multiple user interactions in the beam management procedure, to predict future trajectory data and to predict the best future beam and/or future best subset of beams per UE. A single user class or multiple classes of users (e.g., vehicle class users, pedestrian class users) can be considered for the predicted future trajectory data and or beam prediction data.