Wireless AI/ML Model Deployment Through Control and Data Channels

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing communication systems lack a standardized method for managing the transmission of AI/ML models due to capability limitations of terminal devices, necessitating a flexible and efficient approach for deploying these models.

Innovation Solution

The method involves sending information associated with AI/ML models through control messages or data channels from network devices to terminal devices, allowing deployment under capability constraints, using control and user planes for flexible model deployment.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Adaptability or versatility

If AI/ML models are deployed on terminal devices, then system adaptability and intelligence are improved, but terminal device capability limitations are exceeded

Engineering Contradiction:
Improvesystem adaptabilityVSAvoidterminal device capability
Core Design Contradiction:
Adaptability or versatilityVSReliability

Solution Approach 1:

The patent segments the AI/ML model deployment process into multiple components: model information is divided into model identifiers, configuration parameters, and weight data, which are transmitted separately through control messages and data channels. This segmentation allows the system to manage complex model deployment while respecting terminal device capability limitations by transmitting only necessary portions of model data.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The network device acts as an intermediary between the model source and the terminal device. It manages the storage, processing, and transmission of model information, filtering and preparing data before sending it to terminals. This intermediary role protects terminal devices from directly handling overwhelming model data while still enabling advanced AI functionality.

Inventive Principle:
Principle #24Intermediary (Mediator)

2Adaptability or versatility

If model information is transmitted through control messages, then deployment flexibility is improved, but transmission overhead increases

Engineering Contradiction:
Improvedeployment flexibilityVSAvoidtransmission overhead
Core Design Contradiction:
Adaptability or versatilityVSQuantity of substance

Solution Approach 1:

The patent merges model information transmission with existing control message structures in the communication protocol. Model identifiers, configuration parameters, and weight data are integrated into standard control messages and data channels already present in the system, avoiding the need for separate dedicated transmission channels and reducing overall transmission overhead.

Inventive Principle:
Principle #5Merging (Combining)

Solution Approach 2:

The system transmits only partial model information (model identifiers and configuration parameters) through control messages, while more extensive weight data is transmitted through efficient data channels. This partial action approach balances deployment flexibility with transmission efficiency by sending minimal necessary information through overhead-prone control channels.

Inventive Principle:
Principle #16Partial or excessive action

3Measurement precision

If terminal devices acquire model information through model training processes, then model accuracy is improved, but deployment complexity and time consumption increase

Engineering Contradiction:
Improvemodel accuracyVSAvoiddeployment complexity
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The network device performs preliminary processing of model information before transmission to terminal devices. It pre-processes weight data, extracts essential configuration parameters, and prepares model information in advance, eliminating the need for terminal devices to perform complex model training processes and reducing deployment complexity while maintaining model accuracy.

Inventive Principle:
Principle #10Preliminary action

Solution Approach 2:

Instead of requiring terminal devices to perform full model training, the system transmits pre-trained model copies (model identifiers, configuration parameters, and weight data) from the network device to terminals. This copying approach allows terminals to deploy accurate models without undergoing the complex and time-consuming training process themselves.

Inventive Principle:
Principle #26Copying

Data Source

PatentEP4614919A1Wireless communication method, device, and storage medium
Publication Date: 2025.09.10 GUANGDONG OPPO MOBILE TELECOMMUNICATIONS CORP LTD
  • EP4614919A1 patent drawingFigure 1~4
  • EP4614919A1 patent drawingFigure 5~8
  • EP4614919A1 patent drawingFigure 9~12

AI summary

Embodiments of the present application provide a wireless communication method, a device, and a storage medium. The method comprises: a core network device or an access network device sends first information to a terminal device, the first information comprising information associated with a first model, and a sending mode of the first information comprising: sending the first information by means of a control message and/or sending the first information by means of a data channel.