AI Beam Prediction Using Beam IDs and Network Assistance
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing beam management procedures in wireless communication systems face challenges in efficiently predicting optimal beam pairs due to the limitations of analog beamforming and the dynamic nature of beamforming weights, leading to increased hardware and data collection overheads.
Innovation Solution
A method involving a wireless device (WD) that utilizes an artificial intelligence model to predict beam pairs by measuring a set of beams and requesting assistance information from a network node for beam IDs not included in its training data, allowing the model to be trained and updated dynamically.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If beam management procedures are performed to establish and maintain suitable beam pairs, then coverage and connection reliability are improved, but hardware overhead and data collection requirements increase
Solution Approach 1:
The system performs preliminary beam identification and categorization by comparing beam IDs against stored training data before full beam management procedures. This preliminary action filters out beams that can be handled by existing models, reducing the need for extensive hardware resources and data collection for all beams.
Solution Approach 2:
The patent uses stored training data and beam ID patterns as copies or representations of actual beam characteristics. Instead of collecting and processing complete beam measurement data for every beam, the system uses compact beam ID identifiers and their relationships to training data as simplified copies, reducing hardware overhead.
2Reliability
If beam management procedures are performed to establish and maintain suitable beam pairs, then connection reliability is improved, but data collection overhead increases
Solution Approach 1:
The system extracts only the essential beam identification information (beam IDs) from complete beam measurement data. By separating the identification function from the full measurement data collection, the system reduces data collection overhead while maintaining the ability to establish reliable beam pairs through AI model processing of the extracted IDs.
Solution Approach 2:
The system performs preliminary matching of beam IDs against training data before initiating full beam management procedures. This preliminary action identifies beams that can be handled by existing models, reducing the quantity of new data that needs to be collected and processed.
3Measurement precision
If AI models are trained with comprehensive beam data, then beam prediction accuracy is improved, but training time and computational resources increase
Solution Approach 1:
The patent uses beam ID identifiers and their relationships as simplified copies or representations of complex beam characteristics. The AI model is trained on these compact representations rather than full beam measurement data, maintaining prediction accuracy while significantly reducing training time and computational requirements.
Solution Approach 2:
The system changes the parameters used for AI model training from comprehensive beam measurement data to compact beam ID identifiers and their categorical relationships. This parameter transformation maintains the essential information needed for accurate beam prediction while reducing the dimensionality and computational complexity of the training process.
4Productivity
If beam IDs are used to identify and categorize beams, then data processing efficiency is improved, but system complexity increases
Solution Approach 1:
The system uses beam ID identifiers as simplified copies or representations of complex beam characteristics. These compact IDs enable efficient data processing and AI model processing while the complexity of beam management is handled by the AI system, creating a balanced architecture that improves productivity without proportionally increasing overall system complexity.
Data Source
AI summary
A wireless device (WD) is described. The WD is configured to communicate with a network node and to determine one or more beam identifiers (IDs) one or both of are not included in training data of the artificial intelligence model and have not been used for training an artificial intelligence model. If the one or more beam IDs one or both of are not included in the training data of the artificial intelligence model and have not been used for training the artificial intelligence model, a first request is transmitted to the network node requesting assistance information associated with the one or more beam IDs and/or the assistance information is received. In addition, the WD is configured to cause the artificial intelligence model to be trained using the received assistance information and perform one or more actions using the artificial intelligence model.