Autoencoder Anomaly Detection in Radio Access Networks
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Anomaly detection in Radio Access Networks (RAN) during test and validation phases is time-consuming and inefficient, often requiring manual trial and error to identify issues related to network configuration parameters, which can be costly and labor-intensive, and lacks precision in determining the specific configuration changes needed to resolve anomalies.
Innovation Solution
A method and apparatus using an autoencoder with an encoder and decoder to automate anomaly detection by comparing reconstructed and calculated network performance indicators, identifying deviations, and estimating network configuration parameters to determine if anomalies are related to configuration settings, and selecting optimal parameter values using machine learning to optimize network performance.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If manual trial and error policy is performed on network configuration parameters, then expert intervention can identify anomalies, but the process is time-consuming and lacks precision
Solution Approach 1:
The patent replaces the manual mechanical trial-and-error process with an automated machine learning system. The autoencoder neural network automatically analyzes network configuration parameters and performance indicators to detect anomalies, eliminating the need for expert manual intervention and significantly reducing detection time while maintaining or improving precision.
Solution Approach 2:
The system enables self-service anomaly detection by automatically processing network traces and configuration parameters through the trained autoencoder model. The system independently identifies anomalies and their causes without requiring external expert analysis, making the process autonomous and efficient.
2Reliability
If conventional trial and error policy is used to identify configuration parameter issues, then some anomalies may be resolved, but there is no guarantee that the raised anomalies can be solved and which parameters to change
Solution Approach 1:
The patent implements a feedback mechanism where the autoencoder model continuously learns from network performance data and configuration parameters. The system provides feedback on which specific parameters are causing anomalies and what optimal values should be applied, enabling reliable and systematic parameter adjustment rather than random trial and error.
Solution Approach 2:
The system performs preliminary analysis of network configurations using the trained autoencoder model before actual network deployment or changes. This preliminary action identifies potential anomalies and determines optimal parameter values in advance, ensuring that subsequent parameter adjustments are based on proven effective configurations rather than guesswork.
3Measurement precision
If manual investigation of collected traces is performed, then anomalies can be identified, but the process is time-consuming and long from analysis to feedback
Solution Approach 1:
The patent replaces manual trace investigation with automated machine learning processing. The autoencoder neural network automatically processes network traces and performance indicators at high speed, maintaining the accuracy of anomaly identification while dramatically increasing detection speed and overall productivity.
Solution Approach 2:
The system changes the approach from manual parameter analysis to automated parameter processing by the neural network. The autoencoder model processes multiple parameters simultaneously and identifies anomalies based on learned patterns, achieving both high accuracy and fast processing speeds that manual investigation cannot match.
Data Source
Figure 1~2
Figure 3
Figure 4
AI summary
An apparatus is proposed for anomaly detection in a network, using an autoencoder comprising an encoder and a decoder. The apparatus comprises a processor and a memory including computer program code, causing the apparatus to perform: - providing the decoder with network configuration parameters used to obtain calculated network performance indicators, - obtaining reconstructed network performance indicators from the decoder based on the network configuration parameters used to obtain the calculated network performance indicators, - comparing the reconstructed network performance indicators with the calculated network performance indicators, - detecting an anomaly when observing a deviation between the reconstructed network performance indicators and the calculated network performance indicators, - providing the encoder with the calculated network performance indicators, - obtaining estimated network configuration parameters from the encoder based on the calculated network performance indicators, - detecting that the anomaly is related to the network configuration parameters when observing a deviation between the estimated network configuration parameters and the network configuration parameters used to obtain the calculated network performance indicators.