Wireless Congestion Event Notification for Federated Learning Reporting
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing wireless communication systems face challenges in managing congestion control effectively, leading to delays and skipping of training result reporting during federated learning operations, which are crucial for AI/ML-based network optimization.
Innovation Solution
Implementing a method for setting and notifying congestion control events at the NAS level using AI/ML-based federated learning, involving network functions like AMF and SMF to manage congestion control applications, select terminals for learning iterations, and apply backoff timers to prevent delays and ensure complete training result reporting.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If congestion control is applied to terminals in federated learning operations, then network reliability is improved, but learning iteration speed deteriorates
Solution Approach 1:
The network function sends a notification message to the terminal in advance, informing it of the congestion control application. This allows the terminal to prepare and report training results before congestion occurs, ensuring reliable reporting while maintaining learning iteration speed through proactive coordination
Solution Approach 2:
The system implements a feedback mechanism where the terminal reports training results to the network function, which then notifies the application function. This feedback loop ensures that learning iterations continue smoothly while congestion control is applied, resolving the contradiction between reliability and speed
2Stability of the object's composition
If backoff timer is applied to prevent congestion, then network stability is improved, but training result reporting completeness deteriorates
Solution Approach 1:
The notification message is sent to the terminal before the backoff timer is applied, allowing the terminal to complete and send training results in advance. This preliminary action ensures that training result reporting completeness is maintained while network stability is improved through congestion control
Solution Approach 2:
The system takes preliminary anti-action by notifying the terminal of upcoming congestion control, enabling the terminal to counteract the potential information loss by completing reporting before the backoff timer takes effect, thus preventing training result skipping
Data Source
AI summary
A method of operating a network function in a wireless communication system may comprise receiving a subscription request for a congestion control application related information event for at least one terminal, applying congestion control to the at least one terminal, and performing a notification of a congestion control application related information event based on the applied congestion control. The at least one terminal may be terminals selected based on a federated learning (FL) operation.


