AI/ML Beam Management Adaptation Across Radio Codebooks
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
AI/ML models deployed in cellular radio access networks face challenges in generalizing to unseen data and adapting to changing radio conditions and configurations, leading to degraded performance when trained on specific datasets.
Innovation Solution
A framework for adapting AI/ML models by providing assistance information, training new models, or retraining existing models to enhance their generalization capacity, using techniques like firmware over the air updates and online training processes.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If an AI/ML model is trained with a dataset comprising beams transmitted with a particular codebook, then the model achieves good performance for that specific codebook, but it has difficulties inferencing with beams transmitted with a different codebook
Solution Approach 1:
The patent applies parameter changes by transforming beam measurement parameters through codebook-specific mapping relationships. The network device provides assistance information containing mapping rules that adapt the input parameters to the AI/ML model based on the actual codebook used, allowing the same model to accurately process beams from different codebooks by adjusting the parameter representation rather than changing the model itself
Solution Approach 2:
The patent introduces an intermediary mechanism in the form of assistance information and mapping relationships provided by the network device. This intermediary layer translates between different codebook representations and the standardized input format expected by the AI/ML model, enabling the model to generalize across different codebooks without retraining while maintaining measurement accuracy
2Measurement precision
If an AI/ML model is trained with a dataset representing a certain radio condition environment, then the model performs well under those conditions, but it experiences degraded performance if different channel conditions are met in the field
Solution Approach 1:
The patent addresses radio condition variability by dynamically adjusting input parameters based on current channel conditions. The network device provides assistance information that adapts the beam measurement parameters according to the actual radio environment, allowing the AI/ML model to maintain accurate predictions across diverse channel conditions without requiring separate models for each environment
Solution Approach 2:
The patent implements dynamics by making the AI/ML inference process adaptive to changing radio conditions in real-time. The mapping relationships and assistance information are configured to reflect current network conditions, enabling the model to dynamically adjust its processing to match the prevailing radio environment rather than relying on static training conditions
3Adaptability or versatility
If multiple AI/ML models are stored in the UE to cover different conditions, then the model can adapt to various scenarios, but the device complexity and memory requirements increase
Solution Approach 1:
The patent applies universality by designing a single AI/ML model that can handle multiple different codebooks and radio conditions through parameter adaptation. The assistance information and mapping relationships enable one model to perform the function of what would otherwise require multiple specialized models, reducing UE complexity while maintaining broad adaptability across different scenarios
Data Source
AI summary
An apparatus configured to generate, for transmission to a serving cell, capability information comprising a list of models employing artificial intelligence (AI) or machine learning (ML) for beam management that are currently stored by a user equipment (UE), wherein each model is associated with first conditions for which the model is valid, process, based on signals received from the serving cell, assistance information to adapt beam measurements for a first model stored by the UE to be used under second conditions different from the first conditions for which the first model is valid and adapt the beam measurements based on the assistance information to generate input data for the first model.