AI-Predicted Spatial Filters for Faster NR Beam Management
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The overhead and delay incurred by the traditional beam management process in new radio (NR) systems due to the need to traverse all combinations of transmission and receiving beams during downlink beam sweeping are significant.
Innovation Solution
A wireless communication method utilizing an AI/ML model to predict optimal spatial filters and their link quality based on measurement data, reducing the need for exhaustive beam sweeping by predicting the best beam pairs and their dwelling times.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional beam sweeping is performed by traversing all combinations of transmission and receiving beams, then optimal beam selection can be achieved, but overhead and delay increase significantly
Solution Approach 1:
The system performs preliminary beam measurements and collects measurement data in advance, storing it in a measurement data set. This preliminary action allows the AI/ML model to predict optimal beams for future instances without performing exhaustive measurements each time, thereby reducing delay while maintaining selection accuracy.
Solution Approach 2:
An AI/ML model is introduced as an intermediary between beam measurement and beam selection. The model takes measurement data as input and predicts optimal spatial filters and their dwelling times, eliminating the need for direct exhaustive beam sweeping and reducing the time required for beam management.
2Adaptability or versatility
If traditional beam sweeping traverses all beam combinations, then complete beam coverage is achieved, but signaling overhead increases
Solution Approach 1:
The system extracts only the essential measurement data needed for beam prediction into a compact measurement data set, containing identification information of spatial filters and link quality information. This extraction reduces the amount of data that needs to be processed and signaled, thereby reducing overhead while maintaining the ability to predict optimal beams.
Solution Approach 2:
Instead of performing actual exhaustive beam sweeping for each prediction instance, the system uses the AI/ML model to generate predicted beam information based on historical measurement data. This copying approach allows the system to maintain beam coverage adaptability while significantly reducing the signaling overhead associated with real-time exhaustive measurements.
3Measurement precision
If exhaustive beam measurement is performed in each prediction instance, then accurate beam selection is achieved, but processing complexity increases
Solution Approach 1:
Measurement data is collected and stored in advance in a structured measurement data set during periods when exhaustive measurements are feasible. This preliminary action moves the complex measurement process away from real-time operation, reducing processing complexity during prediction instances while maintaining prediction accuracy through the use of pre-collected data.
Solution Approach 2:
The AI/ML model serves as an intermediary that handles the complex task of beam prediction by processing measurement data and generating predictions for optimal spatial filters and dwelling times. This intermediary approach consolidates the computational complexity into a dedicated model, simplifying the overall beam management process while maintaining prediction accuracy.
Data Source
AI summary
A wireless communication method includes: acquiring, by a first communication device, a first measurement data set; where the first measurement data set includes at least one of: identification information of spatial filters in M measurement instances, or link quality information corresponding to the spatial filters in the M measurement instances; and inputting, by the first communication device, the first measurement data set into a first network model, to output a first prediction data set; where the first prediction data set includes at least one of: identification information of predicted K spatial filters in respective prediction instances of F prediction instances, link quality information corresponding to the predicted K spatial filters in the respective prediction instances of the F prediction instances, or dwelling time of the predicted K spatial filters in the respective prediction instances of the F prediction instances.