Air-Interface AI Model Transmission With Adaptive Delivery Modes
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing communication systems face challenges in determining an appropriate transmission manner for AI models over an air interface, failing to meet the diverse requirements of different application scenarios, which can lead to performance deterioration and assembly failures.
Innovation Solution
A communication method and apparatus that dynamically determines the transmission manner of AI model-related information based on reference information from the receiving device, such as buffering capability and latency requirements, allowing for flexible and block-based transmission to optimize performance.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If a fixed transmission manner is used for AI models, then the transmission process is simple, but it cannot satisfy diverse requirements of different application scenarios leading to performance deterioration
Solution Approach 1:
The patent implements dynamic transmission manner determination by having the first device select different transmission modes (full model transmission, differential model transmission, or parameter transmission) based on real-time reference information from the second device, such as buffering capability and latency requirements, making the system adaptable to varying application scenarios
Solution Approach 2:
The patent employs feedback mechanisms where the second device sends reference information (buffering capability, latency requirements, candidate model sets) to the first device, which then uses this feedback to determine the appropriate transmission manner, ensuring the transmission strategy aligns with the receiving device's capabilities and requirements
2Reliability
If AI model related information is transmitted without considering buffering capability, then the transmission is straightforward, but it causes assembly failures and performance deterioration
Solution Approach 1:
The patent performs preliminary assessment of the second device's buffering capability before transmitting AI model information. The first device receives reference information about buffering capability in advance and uses this information to determine an appropriate transmission manner that ensures the second device can successfully assemble the AI model without overflow or assembly failures
3Loss of time
If transmission latency requirements are not considered, then the transmission process is simple, but it leads to timeout and affects AI model use performance
Solution Approach 1:
The patent dynamically adjusts the transmission manner based on latency requirements received as reference information. For time-sensitive applications, the system selects transmission modes that minimize latency, such as transmitting only essential parameters or using differential models, ensuring the AI model is delivered within the required time frame
Solution Approach 2:
The patent changes transmission parameters (such as model granularity, compression level, and transmission format) based on the latency requirements indicated in the reference information, allowing the system to optimize transmission speed and meet time constraints for different application scenarios
Data Source
AI summary
Embodiments of this application provide a communication method and apparatus. The method is applied to transmission of related information of an artificial intelligence AI model over an air interface. The method includes: A first device and a second device obtain a transmission manner of related information of an AI model, and transmit the related information of the AI model between the first device and the second device in the transmission manner. The first device is an access network device, and the second device is a terminal device; or the first device is a terminal device, and the second device is an access network device. According to the solutions provided in embodiments of this application, an appropriate transmission manner can be obtained for transmitting the related information of the AI model, to satisfy transmission requirements of the AI model in different application scenarios.