Mobile Network Testing via Machine Learning Prediction

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Current mobile network testing methods require extensive drive testing campaigns to gather statistically significant data, consuming significant time and resources, as they need to simulate various user behaviors to assess network quality effectively.

Innovation Solution

The implementation of deep learning models, specifically supervised machine learning using labeled training data, to predict the probability of binary test results, reducing the necessity for multiple tests and enabling faster data collection by training a machine learning model that can map new examples beyond initial training examples.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If drive testing campaigns are run to gather statistically significant data, then measurement precision is improved, but loss of time increases

Engineering Contradiction:
Improvestatistical significanceVSAvoidtesting time
Core Design Contradiction:
Measurement precisionVSLoss of time

Solution Approach 1:

The system performs preliminary actions by training a machine learning model offline using historical drive testing data. Once trained, the model can quickly predict network quality metrics without requiring new extensive drive testing campaigns, thus resolving the contradiction between statistical significance and testing time.

Inventive Principle:
Principle #10Preliminary action

Solution Approach 2:

Instead of performing physical drive testing to gather statistically significant data, the system creates a virtual copy through machine learning predictions. The trained model replicates the function of extensive drive testing by predicting network quality metrics from limited input data, eliminating the need for time-consuming repeated tests.

Inventive Principle:
Principle #26Copying

2Reliability

If multiple test procedures are run to assess different service qualities, then reliability is improved, but productivity deteriorates

Engineering Contradiction:
Improveservice quality assessmentVSAvoidtesting efficiency
Core Design Contradiction:
ReliabilityVSProductivity

Solution Approach 1:

The machine learning model is designed with multi-functionality to predict multiple service quality metrics (drop call rate, data service accessibility, video re-buffering, etc.) simultaneously from a single set of input features. This universal approach maintains comprehensive service quality assessment while eliminating the need to run separate test procedures for each metric.

Inventive Principle:
Principle #6Universality (Multi-functionality)

Solution Approach 2:

The system merges multiple test procedures into a single integrated prediction process. Instead of running separate tests for different service qualities, the machine learning model combines multiple prediction tasks into one unified operation, where a single prediction can provide information about multiple network quality aspects.

Inventive Principle:
Principle #5Merging (Combining)

3Measurement precision

If extensive drive testing campaigns are conducted, then data accuracy is improved, but device complexity increases

Engineering Contradiction:
Improvedata accuracyVSAvoidtesting equipment
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The system replaces the mechanical complexity of extensive drive testing equipment and procedures with a computational machine learning model. Instead of using specialized testing equipment to gather statistically significant data through physical testing, the solution uses software-based predictions that require minimal hardware intervention.

Inventive Principle:
Principle #28Mechanics substitution (Replace mechanical system)

Solution Approach 2:

The machine learning model enables the network testing system to serve itself by automatically predicting service quality metrics without requiring external drive testing campaigns. The trained model can independently assess network quality using readily available network parameters, eliminating the need for complex external testing infrastructure.

Inventive Principle:
Principle #25Self-service

Data Source

PatentEP3525509B1Method and test system for mobile network testing
Publication Date: 2021.03.31 ROHDE & SCHWARZ GMBH & CO KG
  • EP3525509B1 patent drawingFigure 1~2
  • EP3525509B1 patent drawingFigure 3~4
  • EP3525509B1 patent drawingFigure 5

AI summary

A method for mobile network testing is described wherein at least one testing device (12) is used that is configured to be connected to a mobile network. A predefined set of test procedures is run on the at least one testing device (12) in order to obtain binary test results assigned to at least one test parameter. The binary test results of the test procedures are evaluated via a machine learning model. The machine learning model is trained to predict the probability of at least one binary test result based on the binary test results obtained. Further, a test system (10), a method for mobile network testing as well as a prediction system (20) are described.