Beam Measurement Data Set Selection for Faster FR2 Link Prediction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The large downlink resource overhead and measurement delay associated with downlink beam sweeping in wireless communication systems, particularly in Frequency Range 2 (FR2), are significant challenges due to the need for extensive reference signal measurements.
Innovation Solution
A method and apparatus for data set determination that involves a terminal transmitting a data set to a network device for training a first model, and a terminal acquiring a data set for training a second model, utilizing input and label instances based on downlink reference signal measurements to predict target downlink reference signals and link quality, thereby reducing the need for extensive data collection.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If extensive downlink reference signal measurements are performed for beam sweeping, then beam management accuracy is improved, but downlink resource overhead increases and measurement delay increases
Solution Approach 1:
The patent applies preliminary action by pre-training machine learning models with extensive beam measurement data offline, so that during actual operation the terminal only needs to perform limited measurements and use pre-computed model predictions, significantly reducing real-time measurement delay while maintaining beam management accuracy
Solution Approach 2:
The patent uses copying by creating a virtual copy of extensive beam measurement data through machine learning models. Instead of actually measuring all possible beam pairs in real-time, the system copies the relationships between reference signals and beam qualities through trained models, allowing fast prediction without exhaustive measurement
2Measurement precision
If extensive downlink reference signal measurements are performed for beam sweeping, then beam management accuracy is improved, but downlink resource overhead increases
Solution Approach 1:
The patent extracts the essential information needed for beam management from extensive measurement data through machine learning models. By training models on representative samples of beam characteristics, the system extracts the key relationships between reference signals and beam qualities, eliminating the need to transmit and process all possible measurement data, thus reducing downlink resource overhead while maintaining accuracy
Solution Approach 2:
The patent changes the parameter approach from measuring all possible beam pairs exhaustively to measuring a representative sample and using machine learning to infer the rest. This parameter change from complete enumeration to sample-based inference significantly reduces the quantity of downlink resources needed while preserving beam management accuracy
3Measurement precision
If machine learning models are trained using extensive data, then prediction accuracy is improved, but data collection requirements increase
Solution Approach 1:
The patent applies preliminary action by performing extensive data collection and model training in advance during system development or initial setup. The machine learning models are pre-trained with comprehensive beam measurement data before actual operation begins, so that during normal operation the system only needs to collect limited new data for model refinement, significantly reducing ongoing data collection time while maintaining high prediction accuracy
Data Source
AI summary
A data set determination method and apparatus, a communication device, a chip and a storage medium are provided. The method includes: a terminal sends a data set to a network device, the data set includes at least one input instance and at least one tag instance; the input instance is obtained by the terminal measuring a downlink reference signal in a first downlink reference signal set or a second downlink reference signal set, and the tag instance is obtained by the terminal measuring a downlink reference signal in a third downlink reference signal set; and the data set is used for training a first model on the network device and the first model is used for predicting at least one of a target downlink reference signal or target link quality.