Vehicle Component ML Training with Cluster-Balanced Datasets

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

State-of-the-art machine learning methods for vehicle steering systems face inefficiencies due to unevenly distributed datasets, leading to poor performance with previously unknown data, impaired robustness, and high computational overhead, particularly in safety-critical environments.

Innovation Solution

A method involving clustering algorithms to create a uniformly distributed training dataset by focusing on the smallest cluster, ensuring equal representation of scenarios, and using a machine learning algorithm like an artificial neural network with test and validation data to optimize performance across all clusters.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If very large training datasets are used to reflect target behavior, then performance accuracy is improved, but computational overhead and training time increase significantly

Engineering Contradiction:
Improveperformance accuracyVSAvoidtraining time
Core Design Contradiction:
Measurement precisionVSLoss of time

Solution Approach 1:

The patent extracts only the most relevant and informative data points from the large dataset by identifying and removing duplicate scenarios. Instead of using all available data, the system selectively extracts a reduced subset that maintains representative coverage of all driving scenarios, thereby reducing computational overhead while preserving performance accuracy.

Inventive Principle:
Principle #2Taking out (Extraction)

Solution Approach 2:

The patent changes the parameter of dataset size by dynamically determining an optimal reduced size based on scenario diversity rather than using a fixed large size. The system adjusts the training dataset parameters to include only necessary data points, transforming the approach from brute-force large-scale training to optimized selective training.

Inventive Principle:
Principle #35Parameter changes

2Measurement precision

If very large training datasets are used to reflect target behavior, then performance accuracy is improved, but computational resources required for processing increase

Engineering Contradiction:
Improveperformance accuracyVSAvoidcomputational resources
Core Design Contradiction:
Measurement precisionVSUse of energy by moving object

Solution Approach 1:

The system extracts only the essential data points needed for accurate performance by removing redundant duplicate scenarios. This extraction process reduces the computational burden on processing resources while maintaining the accuracy required for safe vehicle operation across all scenarios.

Inventive Principle:
Principle #2Taking out (Extraction)

3Manufacturing precision

If iterative manual dataset generation is performed with trial and error, then dataset quality is improved, but processing time becomes very time-consuming

Engineering Contradiction:
Improvedataset qualityVSAvoiddataset generation speed
Core Design Contradiction:
Manufacturing precisionVSProductivity

Solution Approach 1:

The system performs self-service by automatically identifying duplicate scenarios and determining the optimal reduced dataset composition without requiring iterative manual intervention. The automated algorithm independently evaluates scenario diversity and selects the most representative data points, eliminating the need for time-consuming manual trial and error while maintaining high dataset quality.

Inventive Principle:
Principle #25Self-service

4Reliability

If all data points are considered in the training dataset, then comprehensive coverage is improved, but training efficiency decreases

Engineering Contradiction:
Improvecomprehensive coverageVSAvoidtraining efficiency
Core Design Contradiction:
ReliabilityVSProductivity

Solution Approach 1:

The patent applies local quality by ensuring that each local scenario category is adequately represented in the reduced dataset. Instead of uniform random sampling, the system selectively retains data points that maintain comprehensive coverage of all scenario types, with each local scenario receiving appropriate representation proportional to its importance and diversity.

Inventive Principle:
Principle #3Local quality

Solution Approach 2:

The system changes the parameter of data point selection from including all points to including only a optimized subset. By dynamically adjusting which data points are included based on scenario diversity analysis, the system maintains comprehensive coverage while dramatically improving training efficiency.

Inventive Principle:
Principle #35Parameter changes

Data Source

PatentEP3944158B1Computer-implemented method for machine learning for operating a vehicle component and method for operating a vehicle component
Publication Date: 2025.10.29 VOLKSWAGEN AG
  • EP3944158B1 patent drawingFigure 1
  • EP3944158B1 patent drawingFigure 2
  • EP3944158B1 patent drawingFigure 3

AI summary

The invention relates to a computer-implemented method, in particular for preparing or generating a training dataset, for machine learning for operating a vehicle component (20), as well as a corresponding method for operating a vehicle component (20) and a control unit (10). In a first step, provided multidimensional data points (F1, ..., Fn) are divided into multidimensional clusters (C1, ..., Cm) using a clustering algorithm. Subsequently, a training dataset is generated by selecting data points from the basic training dataset. This selection includes determining the smallest cluster (Ci) among the multitude of clusters (C1, ..., Cm) with the fewest data points. Furthermore, at least a subset of the data points of the smallest cluster (Ci) is provided for the training dataset.In a further step, a subset of data points is selected from each of the remaining clusters (C1, ..., Ci-1, Ci+1, ..., Cm) for the training data set, where the number of selected data points of each remaining cluster (C1, ..., Ci-1, Ci+1, ..., Cm) corresponds to the number of selected data points of the smallest cluster (Ci).