Hybrid Gradient Transmission Using Side Information in Wireless AI
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing wireless communication systems face challenges in efficiently integrating artificial intelligence (AI) into the physical layer due to the need for extensive training data and the mismatch between static training environments and dynamic radio channel characteristics, particularly in deep learning-based AI algorithms.
Innovation Solution
A method and apparatus where user equipment (UE) obtains local gradient information, extracts side information, and transmits it via a digital transmission link to a base station, concentrating on non-common information power and enhancing the reliability of local gradient summation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If deep learning-based AI algorithms are used in wireless communication systems, then communication performance can be improved, but extensive training data is required and the static training environment mismatches with dynamic radio channel characteristics
Solution Approach 1:
The gradient information is segmented into two parts: side information (mean values) transmitted via digital transmission link, and remaining gradient information transmitted via air interface. This segmentation allows the system to benefit from both digital transmission reliability and airComp efficiency without requiring extensive training data for the entire gradient transmission
Solution Approach 2:
Side information (mean values of local gradients) acts as an intermediary that is transmitted through the reliable digital transmission link. This intermediary provides reference information that helps the base station reconstruct the full gradient information more accurately, reducing the need for extensive training data in the deep learning model
2Productivity
If all local gradient information is transmitted via air interface, then transmission efficiency is maintained, but reliability decreases due to channel fading and interference
Solution Approach 1:
The gradient information is divided into side information (transmitted digitally with high reliability) and remaining information (transmitted via air interface with high efficiency). This segmentation allows the critical mean values to be transmitted reliably while maintaining overall transmission efficiency
Solution Approach 2:
The patent merges two transmission approaches: digital transmission for side information and airComp for remaining gradient information. By combining these two methods, the system achieves both reliability (from digital transmission) and efficiency (from air interface transmission) in gradient information delivery
3Loss of information
If mean information common to all elements of local gradient is transmitted, then transmission completeness is improved, but non-common information power is diluted
Solution Approach 1:
The mean information (side information) is extracted from the local gradient and transmitted separately via digital transmission link. This extraction concentrates the non-common information power in the remaining gradient components transmitted via air interface, while still providing complete information through the combination of both transmission paths
Data Source
AI summary
The present specification discloses a method and device. The method, which is for transmitting side information in a wireless communication system and performed by a terminal, comprises the steps of: transmitting a random access (RA) preamble to a base station; receiving a random access response (RAR) from the base station; performing a radio resource control (RRC) connection procedure with the base station; and establishing a digital transmission link with the base station, wherein the terminal acquires local gradient information about at least one resource, acquires side information from the local gradient information, and transmits the side information to the base station via the digital transmission link.