Terminal Radio And Environment Reporting for AI Model Training
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing AI data collection technologies for AI model training in communication systems face challenges in accurately matching network environments due to insufficient adaptation of collected data features, leading to inefficiencies in training and inference accuracy.
Innovation Solution
A communication method that involves a terminal device sending both radio and non-radio information to a network device, where non-radio information provides physical environment details, enabling better matching of training data for AI models, and includes mechanisms for centralized, delayed, or changed reporting to optimize data transmission.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If only radio information is collected for AI training, then data collection is simple, but training data cannot accurately match diverse network environments
Solution Approach 1:
The patent segments the data collection process into two independent parts: radio information (measurement data from terminal devices) and non-radio information (environmental context data). This segmentation allows each type of data to be collected and processed separately, then combined for comprehensive AI training, thereby improving adaptability without creating a single complex collection system
Solution Approach 2:
The patent adds a new dimension to data collection by introducing non-radio information (physical environment details) alongside traditional radio information. This dimensional expansion transforms the data structure from one-dimensional (radio only) to two-dimensional (radio + environment), enabling the AI model to handle diverse network environments more effectively
2Measurement precision
If comprehensive radio and non-radio information is sent to network device, then training data accuracy is improved, but signaling overhead increases
Solution Approach 1:
The patent implements dynamic data reporting mechanisms where the terminal device determines whether to report non-radio information based on triggering conditions configured by the network device. This dynamic approach allows the system to adaptively adjust reporting behavior, sending comprehensive data only when needed to maintain accuracy while reducing signaling overhead during normal operation
Solution Approach 2:
The patent changes the parameter of data reporting from static (always report or never report) to conditional (report based on triggering conditions). By introducing configurable triggering conditions, the system can adjust the reporting behavior to balance between data completeness for training accuracy and signaling overhead reduction
3Reliability
If non-radio information is reported frequently to ensure data freshness, then AI model accuracy is maintained, but signaling overhead increases
Solution Approach 1:
The patent implements periodic reporting mechanisms where the terminal device reports non-radio information at configured intervals or when specific triggering conditions are met. This periodic action ensures data freshness is maintained through regular updates while avoiding continuous signaling by utilizing time-based and condition-based intervals
Data Source
AI summary
This application provides a communication method and apparatus, a readable storage medium, and a chip system, to resolve a problem that accuracy of AI model training, inference, or optimization is not high due to poor adaptation of a training data feature to an AI model. In this application, a terminal device sends radio information to a network device, where the radio information is measurement information obtained by the terminal device (UE); and the terminal device further sends non-radio information to the network device, where the non-radio information is non-radio information corresponding to the radio information, and the non-radio information indicates physical environment information related to the radio information.