AI/ML Model Reporting for Low-Overhead Downlink Beam Prediction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing radio communication technologies in NR downlink beam management require extensive resource consumption and complex UE measurements due to the need for transmitting and measuring CSI-RS or SSB signals on all Tx beams, leading to high overhead and complexity.
Innovation Solution
Employing artificial intelligence (AI) or machine learning (ML) models for downlink beam prediction, where a user equipment (UE) determines and reports first information indicating these models to a network device, allowing for reduced measurements and optimized beam management.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If CSI-RS or SSB signals are transmitted on all Tx beams, then downlink beam management can be performed, but resource consumption increases
Solution Approach 1:
The patent applies partial action by having the UE report only a subset of beam measurement results (e.g., top N beams) rather than all beams. This reduces the uplink resource consumption for reporting while still providing sufficient information for the gNB to perform effective beam management. The UE can be configured to report only the strongest beams or a limited number of beams based on measurement thresholds.
Solution Approach 2:
The patent extracts only the essential beam information needed for downlink beam management from the complete set of beam measurements. By reporting selected beam results rather than all measurements, the system removes unnecessary data transmission overhead while retaining the critical information required for beam selection and management decisions.
2Reliability
If CSI-RS or SSB signals are transmitted on all Tx beams, then downlink beam management can be performed, but measurement overhead increases
Solution Approach 1:
The patent implements partial action by configuring the UE to measure and report only a subset of beams rather than all transmitted beams. The gNB can configure measurement subsets, thresholds, or maximum report quantities to control the overhead. This approach maintains beam management effectiveness while significantly reducing the quantity of measurement data that needs to be processed and reported.
Solution Approach 2:
The patent segments the beam measurement process into multiple phases or groups. Instead of requiring comprehensive measurement of all beams simultaneously, the system can divide beams into subsets measured at different times or with different priorities, reducing the instantaneous measurement overhead and processing burden on the UE.
3Measurement precision
If UE measures CSI-RSs or SSB signals on all Tx beams, then beam selection accuracy can be improved, but UE implementation complexity increases
Solution Approach 1:
The patent applies partial action by enabling the UE to focus measurement efforts on a subset of beams rather than exhaustively measuring all beams. Configuration parameters such as maximum report quantity, measurement thresholds, and beam subsets allow the UE to achieve sufficient measurement precision for beam selection without the complexity of processing all beam measurements.
Solution Approach 2:
The patent utilizes parameter changes by allowing dynamic configuration of measurement and reporting parameters. The gNB can adjust the number of beams to measure, reporting thresholds, and selection criteria based on channel conditions, UE capabilities, and network requirements. This flexibility enables the system to maintain measurement precision while adapting the complexity level to match UE implementation capabilities.
Data Source
Figure 1~2
Figure 3~4
Figure 5~6
AI summary
The present disclosure provides a model information reporting method, device and apparatus, and a storage medium. The method comprises: a terminal determining first information, the first information being used for indicating one or more artificial intelligence or machine learning (AI/ML) models, and the AI/ML models being used for downlink beam prediction; and sending the first information to a network device.