Reinforcement Learning Beam Selection for 5G Uplink Mis-match
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current 5G-NR access nodes face challenges in determining optimal uplink beams due to the mismatch between downlink and uplink channels, leading to inferior link quality and potential call drops, as the best beams derived from average power statistics may differ significantly from single antenna statistics, and existing methods require sweeping all beams, which is resource-intensive and delayed.
Innovation Solution
Implementing a reinforcement learning model that maintains UL RSRP statistics and defines states, actions, and rewards to prioritize UL beams based on DL beams, allowing for efficient identification of optimal UL beams without sweeping all possible beams, using a priority and secondary beam set approach.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If beam sweeping is performed to identify optimal UL beams, then beam selection accuracy is improved, but resource overhead and latency increase
Solution Approach 1:
The system performs preliminary beam sweeping only for a limited set of candidate beams based on DL beam information and stored statistics, rather than sweeping all possible UL beams. This preliminary action identifies promising candidates that are then used for UL transmission, reducing the time and resources required while maintaining adequate selection accuracy.
Solution Approach 2:
The beam selection process is segmented into multiple stages: (1) DL beam measurement and identification, (2) UL beam candidate identification based on DL beam and statistics, (3) Limited UL beam sweeping for candidate beams, and (4) Final UL beam selection. This segmentation allows the system to focus resources on the most promising beam pairs rather than exhaustively searching all possibilities.
2Reliability
If all UL beams are swept to ensure optimal selection, then link quality is improved, but resource overhead increases
Solution Approach 1:
Instead of performing complete beam sweeping for all UL beams, the system performs partial beam sweeping only for a selected subset of candidate UL beams. The candidate selection is based on DL beam correspondence and historical statistics, which allows the system to achieve adequate link quality without the excessive resource overhead of exhaustive sweeping.
Solution Approach 2:
The system changes the parameter of beam sweeping scope from 'all UL beams' to 'candidate UL beams only'. This parameter change is dynamically adjusted based on DL beam measurements and stored statistics, allowing the system to adapt the sweeping extent to current channel conditions and reduce resource overhead while maintaining acceptable link quality.
3Device complexity
If DL beam average power statistics are used to determine best UL beam, then measurement complexity is reduced, but beam selection accuracy deteriorates
Solution Approach 1:
The system introduces an intermediary mechanism - a statistics storage unit that maintains historical beam measurement data. Instead of relying solely on simple DL beam average power statistics, the system queries this intermediary storage to retrieve relevant historical statistics that bridge the gap between simple measurements and accurate beam selection, reducing the need for complex real-time measurements.
Solution Approach 2:
The system implements feedback by storing and utilizing historical beam measurement statistics from previous transmissions. These feedback statistics inform the selection of UL beam candidates and refine the beam sweeping process, allowing the system to improve beam selection accuracy over time without increasing measurement complexity, as the feedback provides guidance based on past performance.
Data Source
Figure 1~2
Figure 3
Figure 4
AI summary
According to an aspect, there is provided an apparatus for the performing the following. The apparatus implements, separately for at least one downlink beam, a reinforcement learning model, where a state defines which of the plurality of uplink beams belong to a priority beam set for uplink reception corresponding to a downlink beam, an action is defined as an addition of a new uplink beam to the priority beam set, a removal of an uplink beam from the priority beam set or doing nothing and a reward is calculated based on a change in uplink signal-to-noise ratio due to an action adjusted with a cost for taking the action. The apparatus calculates iteratively at least one optimal state using at least one reinforcement learning model based on uplink signal-to-noise ratio statistics and on the plurality of optimal downlink beams for transmission to said plurality of terminal devices.