Trajectory-Based Sector Radio Link Adaptation for Reduced CSI Feedback
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current link adaptation control loops in wireless communication systems, such as those defined by the 3GPP standards, rely heavily on frequent and resource-intensive channel feedback measurements, which are inefficient and costly, especially in narrow beam radio links, due to the unpredictable nature of radio channels.
Innovation Solution
Implementing a reinforcement learning system that predicts and adapts to mobile radio link characteristics by learning the trajectories of wireless devices, reducing the need for frequent CSI measurements through trajectory-based link adaptation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If frequent channel feedback measurements are performed to accurately determine radio link state, then link adaptation accuracy is improved, but radio resource consumption increases
Solution Approach 1:
The system performs preliminary actions by learning and storing radio link characteristics patterns from historical data before actual link adaptation decisions are needed. This pre-learning phase captures channel behavior patterns during periods when resources are available, enabling faster and more accurate decisions during transmission without requiring frequent measurements
Solution Approach 2:
Instead of directly measuring the current radio link state frequently, the system creates copies of historical radio link characteristics and uses these learned patterns to predict current state. The machine learning model replicates the behavior of frequent measurements by generating predicted CQI, PMI, and RI values based on learned trajectories, reducing the need for actual measurements
2Speed
If tight feedback loops are used to track radio link state changes, then link adaptation responsiveness is improved, but interference increases
Solution Approach 1:
The system performs preliminary learning of radio link characteristics patterns during periods when the channel is relatively stable, capturing the temporal and spatial patterns of channel behavior. This pre-acquired knowledge enables the system to predict future channel states without requiring tight feedback loops, thereby reducing interference while maintaining responsiveness
3Reliability
If frequent CSI measurements are performed to maintain accurate link adaptation, then transmission reliability is improved, but battery life decreases
Solution Approach 1:
The system creates copies of historical channel state information and uses machine learning models to generate predicted CSI values based on learned radio link characteristics. This approach maintains transmission reliability by providing accurate CQI, PMI, and RI predictions without requiring the wireless device to perform frequent measurements, thereby conserving battery life
Solution Approach 2:
The network side performs the heavy lifting of learning and predicting radio link characteristics, serving itself with the computational burden. This relieves the wireless device from frequent measurements and processing, allowing the device to conserve energy while the network maintains accurate link adaptation through its learned models
Data Source
Figure 1
Figure 2
Figure 3
AI summary
A method and network node for predicting and adapting to mobile radio link characteristics in a sector are disclosed. According to one aspect, a method includes learning a set of at least one trajectory based at least in part on a first set of observations received from at least one wireless device (WD), a trajectory including a subset of the first set of observations. The method also includes assigning a trajectory in the set of at least one trajectory to a first WD of the at least one WD based at least in part on a second set of observations received from the first WD subsequent to receipt of the first set of observations. The method further includes adjusting an update period for receiving future observations from the first WD based at least in part on the assigned trajectory.