RRC AI Model Transfer Using Split Coding and Selective Retransmission

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing wireless communication systems face challenges in efficiently transmitting control channel information and artificial intelligence model information with varying reliability and latency requirements, particularly in 5G networks, where URLLC and eMBB use cases impose different demands on RAN resource loads.

Innovation Solution

A method is introduced where control channel information and AI model information are transmitted using different encoding formats within a radio resource control signal message, with control channel information being transmitted at a higher reliability and AI model information at a lower reliability, allowing for quicker delivery but with potential retransmission of segments.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Adaptability or versatility

If control channel information and AI model information are transmitted using the same encoding format, then the transmission structure is simple, but the reliability and latency requirements of different information types cannot be met simultaneously

Engineering Contradiction:
Improveability to meet different reliability and latency requirementsVSAvoidtransmission structure complexity
Core Design Contradiction:
Adaptability or versatilityVSDevice complexity

Solution Approach 1:

The RRC message is segmented into distinct portions: a first portion for control channel information and a second portion for AI model information. Each portion is encoded separately with different encoding formats and rates, allowing independent optimization for their respective reliability and latency requirements while maintaining overall message structure.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

Different encoding formats and rates are applied to different portions of the same RRC message based on their specific requirements. The control channel information portion uses a first encoding format with higher reliability, while the AI model information portion uses a second encoding format with lower latency, achieving local optimization within the unified message structure.

Inventive Principle:
Principle #3Local quality

2Reliability

If control channel information is transmitted with high reliability using a lower rate, then the transmission is more reliable, but the transmission time increases

Engineering Contradiction:
Improvecontrol channel information reliabilityVSAvoidtransmission time
Core Design Contradiction:
ReliabilityVSLoss of time

Solution Approach 1:

The message is divided into segments that can be transmitted and decoded independently. The control channel information segment is transmitted with higher reliability encoding, while the AI model information segment uses faster encoding, allowing parallel processing and reducing overall transmission time while maintaining control channel reliability.

Inventive Principle:
Principle #1Segmentation

3Productivity

If AI model information is transmitted quickly using a higher rate, then the transmission speed increases, but the reliability decreases and retransmission may be needed

Engineering Contradiction:
Improvetransmission speedVSAvoidAI model information reliability
Core Design Contradiction:
ProductivityVSReliability

Solution Approach 1:

The system implements feedback mechanisms where the receiving device can request retransmission of specific segments of AI model information if decoding fails. This allows the system to maintain high transmission rates while providing a safety net for reliability through selective retransmission based on feedback about received segment quality.

Inventive Principle:
Principle #23Feedback

4Reliability

If separate encoding formats are used for control channel information and AI model information, then different reliability and latency requirements are met, but the encoding complexity increases

Engineering Contradiction:
Improvedifferential reliability meetingVSAvoidencoding complexity
Core Design Contradiction:
ReliabilityVSDevice complexity

Solution Approach 1:

The RRC message is segmented into distinct portions with separate encoding formats. The first portion for control channel information uses a first encoding format optimized for reliability, while the second portion for AI model information uses a second encoding format optimized for speed. This segmentation allows independent encoding optimization without requiring complex integrated encoding schemes.

Inventive Principle:
Principle #1Segmentation

Data Source

PatentUS12563572B2Artificial intelligence model delivery via radio control plane
Publication Date: 2026.02.24 DELL PROD LP
  • US12563572B2 patent drawing
  • US12563572B2 patent drawing
  • US12563572B2 patent drawing

AI summary

A radio access network node may determine to update an artificial intelligence machine learning model at a user equipment. The node may transmit the model in a radio resource control message that comprises a primary portion and a secondary portion. The primary portion may be used to transmit control information. The secondary portion may be used to transmit the model, or other data. The control message may comprise a format indication indicative of a code rate to be used by the user equipment to decode the primary portion. The radio resource control message may comprise a secondary format indication indicative of a code rate to be used to decode the secondary portion. A format indication may comprise a retransmission indication indicative of retransmission of one or more segments of either, or both, of the primary portion or secondary portion being enabled.