AI Scheduling Configuration Threshold Filtering for Wireless Stability
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing wireless communication systems face performance degradation due to adverse actions resulting from optimized AI configuration values for radio resource scheduling, leading to sudden deterioration in network performance.
Innovation Solution
An electronic device that obtains environmental information from a radio access network, identifies configuration values using a learning model, compares them with threshold values, and adjusts these values to prevent adverse impacts, thereby stabilizing network performance by inputting the adjusted values back into the learning model.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If AI reinforcement learning is used to optimize scheduling configuration values, then scheduling performance is improved, but network stability deteriorates due to sudden performance degradation from adverse actions
Solution Approach 1:
The patent applies preliminary anti-action by introducing a threshold comparison mechanism that prevents the reinforcement learning agent from executing configuration values before they can cause adverse effects. The threshold unit compares proposed configuration values against historical data and prevents application of values that would result in performance degradation, thereby counteracting the inherent instability of RL-based optimization while preserving its scheduling performance benefits
Solution Approach 2:
The patent implements feedback through a closed-loop system where performance data is continuously collected, stored, and used to adjust future configuration value proposals. The threshold unit receives performance information and uses it to determine whether to allow or block configuration changes, creating a feedback mechanism that maintains network stability while enabling AI-driven optimization
2Productivity
If AI reinforcement learning generates configuration values for radio resource scheduling, then optimization is achieved, but harmful adverse actions are performed that degrade network performance
Solution Approach 1:
The patent converts the potentially harmful exploration behavior of reinforcement learning into a beneficial process by using the same RL agent to generate configuration values while simultaneously using a threshold comparison mechanism to filter out harmful actions. The system leverages RL's optimization capabilities while preventing its harmful exploration from degrading network performance, effectively turning a harmful factor into a beneficial controlled process
Solution Approach 2:
The threshold unit serves as an intermediary between the reinforcement learning agent and the actual configuration application. It mediates the interaction by comparing proposed configuration values against historical performance data and blocking harmful actions while allowing beneficial optimizations to be applied, thus preventing adverse effects without completely stopping the optimization process
Data Source
AI summary
An electronic device may include a storage device and at least one processor, wherein the at least one processor may obtain environmental information from a radio access network (RAN) to store the environmental information in the storage device, identify at least one first configuration value for scheduling a radio resource from the obtained environmental information, based on a learning model generated based on previously obtained environmental information, compare the first configuration value with at least one threshold value, adjust the first configuration value to a second configuration value, and transmit the adjusted second configuration value to the RAN.