Network AI Model Switching for Inference Performance Deterioration
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing AI models in communication networks deteriorate in performance due to environmental factors, necessitating a method to quickly identify and address this deterioration to maintain network performance.
Innovation Solution
A communication method and apparatus that monitors AI model performance using configuration information and performance indicators, allowing timely adjustments such as switching models or falling back to non-AI modes when deterioration is detected.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If AI model is deployed in terminal device for inference, then network performance and user experience are improved, but inference performance deteriorates due to radio/non-radio environment factors
Solution Approach 1:
The patent implements a feedback mechanism where the network device monitors performance indicators (such as throughput, block error rate, channel state information) of the AI model deployed in the terminal device. When the monitored performance indicator falls outside a preset range, the network device sends first indication information to trigger model switching, and second indication information to switch back to the original AI model, thereby maintaining reliable inference performance despite environmental variations
Solution Approach 2:
The patent changes the parameters of the AI model by switching between different AI models based on monitored performance indicators. The system selects from multiple AI models with different characteristics (e.g., different accuracy-speed tradeoffs) to adapt to changing radio/non-radio environment factors, thus maintaining both high network performance and reliable inference performance
2Reliability
If AI model switching is implemented to address performance deterioration, then inference performance is maintained, but device complexity increases
Solution Approach 1:
The patent introduces the network device as an intermediary that manages AI model switching. Instead of the terminal device independently managing multiple AI models and switching logic (which would increase terminal complexity), the network device monitors performance indicators and sends indication information to control model switching, thereby maintaining inference performance while reducing terminal device complexity
Solution Approach 2:
The system implements self-service through automated performance monitoring and model switching based on preset ranges. The network device automatically monitors performance indicators, compares them against thresholds, and triggers model switching without manual intervention, reducing operational complexity while maintaining reliable inference performance
Data Source
AI summary
A communication method and apparatus are provided, to identify that inference performance of an artificial intelligence (AI) model in a communication network deteriorates. The method includes: A second network element sends configuration information to a first network element. Correspondingly, the first network element receives the configuration information from the second network element. The configuration information includes a performance indicator of a first AI model and a preset condition corresponding to the performance indicator. The first network element changes the first AI model when a value of the performance indicator of the first AI model meets a preset condition. The first network element can be a distributed unit, the second network element can be a central unit, and the distributed unit and the central unit can be connected through an F1 interface.