Wide Beam Codebook Prediction for Low-Latency Beam Refinement
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The existing beam management procedures in wireless communication systems, such as those in LTE and 5G NR, face challenges with increased CSI-RS measurements and feedback overhead, leading to higher latency and inefficiencies in beam selection and refinement due to the use of high-dimensional MIMO arrays.
Innovation Solution
A method involving the generation of a wide beam codebook based on a grid of narrow beams, combined using a machine learned model, to predict refined beams for improved beam selection and reduce latency by optimizing beam scanning operations.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If traditional beam sweeping and feedback procedures are used for beam management, then beam establishment and maintenance can be achieved, but overhead and latency increase significantly
Solution Approach 1:
The system performs preliminary beam training by having the UE measure all beams in the codebook and report preferred beams in advance. The gNB then predicts future best beams based on these historical measurements and reported preferences, eliminating the need for continuous real-time beam sweeping and feedback during normal operation.
Solution Approach 2:
Instead of performing actual beam sweeping and measurements continuously, the system uses copied historical measurement data and reported beam preferences to predict future beam states. The machine learning model creates virtual beam predictions based on past patterns rather than requiring real-time physical measurements.
2Measurement precision
If frequent beam sweeping is performed to maintain beam accuracy, then beam prediction accuracy is improved, but system overhead increases
Solution Approach 1:
The UE performs comprehensive beam measurements and reports preferred beams preliminarily during initial configuration. After this preliminary measurement phase, the system relies on predictions based on historical data rather than continuous measurements, dramatically reducing ongoing signaling overhead while maintaining prediction accuracy.
Solution Approach 2:
The machine learning model serves itself by continuously learning from historical measurement data and reported beam preferences. The model improves its prediction accuracy autonomously over time without requiring additional measurement overhead, as it extracts patterns from existing data and applies them to predict future beam states.
3Measurement precision
If narrow beams are used for precise beamforming, then beam directionality is improved, but the complexity of beam management increases
Solution Approach 1:
The system uses wide beams as simplified copies or representations of multiple narrow beam directions. Instead of managing numerous individual narrow beams, the gNB transmits wide beams that cover multiple spatial directions, and the UE reports preferences among these wide beams. The ML model then predicts which narrow beam within the wide beam coverage would be optimal, reducing management complexity while maintaining precision.
Data Source
AI summary
Various techniques are provided for determining, by a base station, a beam coverage in a cell, the beam coverage including angle range for an azimuth and an angle range for an elevation, determining, by the base station, an angle quantization step, generating, by the base station, a grid of narrow beams based on the angle range and the angle quantization step, computing, by the base station, a beam index matrix based on the grid of narrow beams, generating, by the base station, a wide beam codebook based on the grid of narrow beams, and transmitting, by the base station, at least one of a CSI-RS resource and a SSB resource using multiple-TX-directions wide beams based on the wide beam codebook.