Scenario Library Clustering for Automated Driving Test Coverage
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for creating or updating scenarios libraries for testing highly-automated driving functions struggle to ensure optimal test coverage of the operational design domain (ODD) taxonomy with efficient resource utilization, as it is difficult to determine the relevance and value of new scenarios for inclusion in the library.
Innovation Solution
A computer-implemented method for criterion-based updating of scenarios libraries, which involves comparing new test scenario data sets to existing data sets and a requirements profile using clustering algorithms, determining degrees of correspondence, and adding or discarding scenarios based on threshold values to ensure optimal test coverage and resource efficiency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Quantity of substance
If new test scenario data sets are added to the scenarios library without systematic evaluation, then the quantity of test scenarios increases, but the test coverage efficiency and resource utilization deteriorate
Solution Approach 1:
The system implements feedback by calculating similarity metrics between new scenario data sets and existing clusters, then using this feedback to determine whether to add the new scenarios. The similarity calculation provides quantitative feedback that guides the decision-making process, ensuring that only scenarios that add meaningful value are included in the library.
Solution Approach 2:
The system changes parameters by transforming scenario data into clustered representations with calculated similarity metrics. By converting scenarios into comparable parameter sets (clusters with similarity scores), the system enables systematic evaluation and selection of new scenarios based on quantitative parameters rather than qualitative assessment.
2Measurement precision
If comprehensive clustering and comparison of scenario elements is performed, then the test coverage accuracy improves, but the computational complexity and processing time increase
Solution Approach 1:
The system segments scenario data into distinct clusters of scenario elements. By dividing the comprehensive scenario comparison into smaller cluster units, the system reduces computational complexity while maintaining measurement precision. Each cluster represents a grouped set of similar scenario elements, making the overall comparison process more manageable.
Solution Approach 2:
The system performs preliminary clustering of scenario elements before conducting the final comparison and evaluation. This preliminary action organizes the data structure in advance, creating ready-to-compare clusters that streamline the subsequent similarity calculation and decision-making processes, reducing overall computational burden.
3Reliability
If the scenarios library includes all possible test scenarios, then the test coverage completeness improves, but the storage requirements and processing overhead increase
Solution Approach 1:
The system merges similar scenario elements into unified clusters. By combining redundant or highly similar scenarios into single cluster representations, the system maintains comprehensive test coverage while significantly reducing the overall library size. The clustering process identifies and merges duplicate or near-duplicate scenarios.
Solution Approach 2:
The clustered scenario elements serve multiple functions: they represent individual test scenarios, group similar scenarios together, and provide a basis for similarity comparison. This multi-functionality allows the system to maintain comprehensive coverage with a more compact representation, as each cluster can fulfill multiple roles in the testing framework.
Data Source
AI summary
A computer-implemented method for criterion-based updating of a scenarios library having virtual vehicle environments for testing automated driving functions of a motor vehicle includes: providing a scenarios library having a number of test scenario data sets and a requirements profile having at least one scenario element for creating or updating the scenarios library; comparing a further test scenario data set having a plurality of scenario elements to at least one cluster of the number of test scenario data sets comprised in the scenarios library and to the requirements profile; and adding the further test scenario data set to the scenarios library or discarding the further test scenario data set.