RRC AI/ML Model Transfer for UE-Base Station Alignment
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Coordination and alignment of AI/ML models between wireless communication devices, such as UEs and base stations, is challenging due to data privacy concerns and vendor-specific model management, leading to difficulties in synchronizing understanding and training across the air interface.
Innovation Solution
Implementing RRC procedures for wireless AI/ML that enable the transfer and alignment of AI/ML models through uplink and downlink RRC messages, allowing for training and inference alignment between UEs and base stations using methods like federated learning and model transfer, with support for one-sided and two-sided models.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If AI/ML models are transferred and trained between UEs and base stations, then model alignment and synchronized understanding are improved, but data privacy concerns and vendor-specific model management complexity increase
Solution Approach 1:
The patent introduces RRC messages as an intermediary mechanism to facilitate secure and standardized AI/ML model transfer between UEs and base stations. The RRC layer acts as a mediator that handles model transmission, configuration, and coordination, thereby improving model alignment while managing the complexity through a dedicated communication interface rather than direct peer-to-peer model management.
Solution Approach 2:
The patent segments the AI/ML model management process into distinct phases handled by specific RRC procedures: model transfer phase, training phase, and inference phase. This segmentation allows each phase to be managed independently with appropriate protocols, reducing overall system complexity while ensuring reliable model alignment across the air interface.
2Reliability
If federated learning and model transfer methods are implemented, then training and inference alignment is improved, but communication overhead and air interface complexity increase
Solution Approach 1:
The patent makes the RRC message structure universal and multi-functional to handle various AI/ML operations including federated learning, model transfer, training coordination, and inference management. By designing a flexible RRC message framework that can accommodate multiple AI/ML protocols and methods, the air interface complexity is reduced while maintaining reliable training and inference alignment across different scenarios.
3Loss of information
If AI models are transmitted via RRC messages, then model synchronization between UE and base station is improved, but signaling overhead and communication resources are consumed
Solution Approach 1:
The patent implements preliminary action by establishing RRC connections and configuring AI/ML model parameters before actual model transmission and training operations. This preliminary setup includes negotiating model formats, agreeing on training parameters, and pre-configuring resource allocation, which reduces the signaling overhead during the actual model synchronization phase while ensuring complete and accurate model transfer.
Data Source
AI summary
There is provided an apparatus of a user equipment (UE), the apparatus comprising a processor, and a memory storing instructions that, when executed by the processor, configure the apparatus to receive an Artificial Intelligence (AI) model from a base station; obtain a trained AI model resulting from a training of the AI model; and send the trained AI model to the base station via an uplink Radio Resource Control (RRC) message.