Wireless Cell On-Off Control Using Graph Neural Networks
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing cellular communication networks face inefficiencies in energy consumption due to over-provisioned cell deployments, leading to unnecessary energy expenditure and potential network performance degradation, especially in dense and future 6G networks with upper mid-band carriers.
Innovation Solution
Implementing an AI-based cell activation/deactivation policy using a Graph Neural Network (GNN) architecture and Graph Reinforcement Learning (GRL) to dynamically manage cell activation and deactivation based on spatial traffic load and network resource interdependencies across neighboring cells, optimizing energy savings while maintaining service quality.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If cells are over-provisioned to meet peak traffic demand, then service quality is improved, but energy consumption increases
Solution Approach 1:
The patent implements dynamic cell activation and deactivation based on real-time traffic conditions. The system transitions from a static cell configuration to a dynamic one where cells are activated only when needed and deactivated when traffic demand is low, resolving the contradiction between maintaining service quality and reducing energy consumption.
Solution Approach 2:
The system changes the operational state parameter of cells between active and inactive states based on traffic load thresholds. By dynamically adjusting this parameter, the system can reduce energy consumption during low-demand periods while ensuring service quality is maintained during peak periods.
2Use of energy by moving object
If cells are deactivated to save energy, then energy efficiency is improved, but network performance may degrade
Solution Approach 1:
The patent employs a feedback mechanism that continuously monitors traffic load on deactivated cells and their impact on neighboring active cells. When performance degradation is detected, the system activates additional cells to maintain service quality, thus resolving the contradiction between energy efficiency and network performance.
Solution Approach 2:
The system performs preliminary evaluation of traffic patterns and cell interdependencies before deactivating cells. By using Graph Neural Networks to predict future traffic demands and assess the impact of cell deactivation, the system can make informed decisions that prevent performance degradation while maximizing energy savings.
3Device complexity
If cell activation decisions are made independently without considering neighboring cells, then device complexity is reduced, but spatial load dependencies are not optimized
Solution Approach 1:
The patent merges the decision-making processes of neighboring cells by using Graph Neural Networks to model spatial dependencies. Instead of independent cell control, the system combines information from multiple cells to make coordinated activation decisions, optimizing energy efficiency while accounting for load distribution across the network.
Data Source
AI summary
Systems, apparatus, articles of manufacture, and methods are disclosed control on-off switching in wireless networks. An example computer readable medium comprises instructions that, when executed, cause at least one programmable circuitry to at least generate a mathematical graph that includes measurement information about a first communication cell and a second communication cell; process a first embedding of the mathematical graph with a graph convolutional network layer to generate an updated embedding; concatenate the first embedding and the updated embedding to generate a concatenated embedding; and process the concatenated embedding via a neural network; and cause the first communication cell to be deactivated based on the result of the neural network.