RAN Slice Feasibility Checks Using ML Resource Demand Estimation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for network slice feasibility checks in radio access networks are time-consuming, less accurate, and dependent on human expertise, lacking flexibility and considering the dynamic nature of network resources.
Innovation Solution
A method utilizing supervised machine learning models to estimate resource demand for new RAN slices by comparing historical data with defined thresholds, incorporating data from network nodes to automate the feasibility check process.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If manual feasibility check methods are used, then human expertise can be applied to evaluate RAN slice requirements, but the process becomes time-consuming and less accurate
Solution Approach 1:
The patent replaces manual expert assessment with automated machine learning models that process historical network data and slice requirements to determine feasibility. The ML-based system eliminates human intervention in the evaluation process, providing consistent, accurate results without time constraints associated with manual analysis.
Solution Approach 2:
The system enables self-service feasibility checking by automatically comparing slice requirements against historical network performance data and resource availability. The automated ML models independently evaluate whether new slices can be accommodated without requiring human expertise or manual resource assessment.
2Adaptability or versatility
If traditional feasibility check methods are used, then existing network knowledge can be leveraged, but the system lacks flexibility to adapt to dynamic network conditions
Solution Approach 1:
The patent implements dynamic adaptability by training machine learning models on historical network data that captures varying traffic patterns, resource utilization, and network conditions. The system continuously learns from new data, adapting to changing network dynamics while maintaining reliable feasibility assessments through statistically robust model training.
Solution Approach 2:
The system performs preliminary actions by pre-training ML models with extensive historical network data before actual feasibility checks. This preliminary training phase enables the system to quickly adapt to new slice requirements and dynamic conditions during operation, maintaining both flexibility and reliability.
3Productivity
If expert-dependent methods are used, then human judgment can be applied to complex scenarios, but operational costs increase
Solution Approach 1:
The patent replaces complex human expert judgment with standardized machine learning models that automatically process slice requirements and network data. This substitution eliminates the need for specialized human expertise in each feasibility check while maintaining high operational efficiency through automated, consistent evaluation procedures.
Solution Approach 2:
The system transforms complex qualitative expert judgment into quantitative parameter-based ML model inputs and outputs. By converting expert knowledge into trainable model parameters and automated decision rules, the system achieves high productivity through standardized processing while reducing operational complexity through algorithmic consistency.
Data Source
AI summary
A method for a feasibility check of a new RAN slice performed by a network node is provided. The method includes computing a first estimate of occupied resources for each resource within at least one of the network node and a cell of the network node. The first estimate performed for each resource using historical data of a measurement of utilization of each resource. The method further includes applying a trained supervised machine learning model to provide a second estimate of resource demand for each resource for the new RAN slice. The method further includes comparing a sum of the first estimate and the second estimate to a defined threshold value; and admitting the new RAN slice when the sum has a value not greater than the defined threshold value per resource.