Radio Resource Management for Network Slices
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current radio resource management in network slicing for radio access networks faces challenges in dynamically adapting to changes in network slice usage without requiring re-training of neural networks for existing slices, leading to delays and inefficiencies when adding or removing slices.
Innovation Solution
Implementing a dynamic slice-aware radio resource management process using reinforcement learning-based neural networks to determine slice-specific cost indices, allowing for independent updates and resource allocation decisions for each network slice without affecting other slices, thus enabling seamless addition or removal of slices without re-training.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If traditional radio resource management is used in network slicing, then resource allocation can be performed, but adding or removing network slices requires re-training of neural networks for existing slices, causing delays and inefficiencies
Solution Approach 1:
The patent divides the radio resource management system into independent network slice-specific processes, where each network slice has its own dedicated process that operates autonomously. This segmentation allows individual slices to be added or removed without affecting other slices, eliminating the need for system-wide re-training and resolving the contradiction between adaptability and time loss.
2Manufacturing precision
If neural networks are re-trained for existing slices when adding or removing slices, then resource allocation accuracy can be maintained, but computational complexity and processing time increase significantly
Solution Approach 1:
The patent implements dynamic resource allocation where each network slice process continuously adapts to changing conditions in real-time without requiring periodic re-training. The slice-specific processes dynamically adjust resource allocation based on current network state, maintaining accuracy while avoiding the computational burden of re-training.
Solution Approach 2:
Each network slice process autonomously manages its own resource allocation decisions using its dedicated neural network, making independent decisions without requiring centralized re-training operations. This self-service approach maintains allocation accuracy while significantly reducing overall system complexity.
3Stability of the object's composition
If centralized radio resource management is used, then coordination across slices can be achieved, but the system cannot scale efficiently with changing network dynamics
Solution Approach 1:
The patent segments the centralized management function into distributed slice-specific processes, where each process independently manages its designated network slice. This segmentation enables the system to scale efficiently by simply adding or removing individual slice processes without impacting the stability or coordination of other slices.
Data Source
Figure 1
Figure 2A~2B
Figure 3
AI summary
According to an example aspect of the present invention, there is provided an apparatus configured to obtain, from user equipment-level operating statistics from a radio access network, network slice-level operating statistics concerning plural network slices in the radio access network, update, using a plurality of processes, each process specific to a distinct network slice, network slice specific cost indices based at least in part on the network slice-level operating statistics, each cost index indicating a relative resource cost of increasing a radio resource allocation of a respective network slice, each process running a distinct neural network to update the respective cost index, determine, based on the cost indices, radio resource configurations for the plural network slices, and control the radio access network to provide radio resources to the plural network slices according to the determined radio resource configurations.