Wireless Network Node ML Coordination for Mobility and Load Balancing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing wireless communication systems face challenges in optimizing network performance using artificial intelligence/machine learning techniques, particularly in 3GPP LTE and NR systems, due to limitations in supporting machine learning models across network nodes and user equipment.
Innovation Solution
A method and apparatus are introduced to facilitate the use of machine learning models by enabling network nodes to exchange capability information, request and process input data, and perform model processing to obtain output data for actions such as mobility optimization, load balancing, and network energy saving, involving transceivers, processors, and memory units.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If machine learning models are deployed in wireless communication systems for optimization, then network performance and service quality are improved, but device complexity and implementation difficulty increase
Solution Approach 1:
The system divides the ML model into multiple segments: some network nodes execute the full model while others execute only parts of it. This segmentation allows complex ML functionality to be distributed across multiple nodes, reducing the complexity burden on individual devices while maintaining overall network performance benefits.
Solution Approach 2:
The patent creates a universal ML execution framework where network nodes can perform multiple functions: acting as ML clients, servers, or both simultaneously. This multi-functionality allows the same infrastructure to support various optimization scenarios (energy saving, load balancing, mobility) without requiring separate dedicated systems for each application.
2Reliability
If machine learning models are used for network optimization, then service quality and user experience are enhanced, but power consumption and operational costs increase
Solution Approach 1:
The system implements self-service through automated ML model execution and optimization without requiring manual intervention. Network nodes autonomously perform model inference, generate optimization decisions, and adapt to changing conditions, reducing the need for human operators and associated operational costs while maintaining high service quality.
Solution Approach 2:
The patent dynamically adjusts ML model parameters and execution frequency based on network conditions and node capabilities. By changing parameters such as model complexity, update intervals, and execution timing, the system optimizes the balance between service quality improvements and power consumption, ensuring ML is used most efficiently.
3Adaptability or versatility
If machine learning capability is supported across all network nodes, then adaptability and optimization potential are improved, but device complexity and implementation difficulty worsen
Solution Approach 1:
The system implements dynamic adaptability where network nodes can change their ML participation status based on current conditions. Nodes can dynamically switch between being ML clients, servers, or neither, allowing the network to adapt its ML capabilities to match available resources, node capabilities, and optimization needs without requiring all nodes to permanently support ML functions.
Solution Approach 2:
The patent introduces intermediary mechanisms for capability advertisement and model distribution. Instead of requiring direct ML implementation at every node, intermediary nodes or centralized controllers facilitate model sharing and coordination, enabling adaptability across the network while reducing the complexity burden on individual participating nodes.
Data Source
AI summary
The present specification relates to machine learning (ML)-based optimization. A method carried out by a first network node in a wireless communication system, according to one embodiment of the present disclosure, comprises the steps of: receiving, from a second network node, a first message including capability information indicating whether the second network node is capable of supporting a ML model; on the basis that the capability information indicates that the second network node is capable of supporting the ML model, transmitting, to the second network node, a second message for requesting for input data for model processing generated by the ML model; receiving the input data from the second network node; carrying out the model processing on the basis of at least one among the input data received from the second network node or a measurement result received from a user equipment (UE); acquiring output data inferred from the model processing; and carrying out one or more operations on the basis of the output data.