Automotive Industry Insights

Potential Implications of Taxonomy-, Labeling-, and Ontology-Based Standards for Testing

Amongst others, this subsection considers:

  • OpenXOntology
  • OpenLABEL
  • ISO 34501
  • ISO 34503
  • ISO 34504

An underlying requirement for collaboration and data exchange is a shared understanding of the terminology being used. Standardized taxonomies such as those mentioned above provide this baseline in the domain (or a subset of it). The overlap, differences, and relationships amongst the terms being used need to be clear for exchange to be successful. Worth emphasizing is that it is not necessary to have only one taxonomy as long as the overlaps and differences in the terms are clear. OpenXOntology currently does this, for example, by clearly mapping terms across the OpenX standards, rather than defining one single taxonomy.

We also recommend the development of an application-specific layer to the OpenXOntology standard for testing that takes all of the conclusions made in this report into consideration. The natural-language-like reasoning enabled by an ontology could also be used to define another layer to scenario and test descriptions that is natural-language-like. This would aim to address, for example, the functional scenario layer when applied to scenario descriptions. Such a layer would enable machine-readable, natural-language-like descriptions in legislature that can easily be converted to more concrete descriptions, thus being less subject to ambiguity or misinterpretation.
The aforementioned taxonomies enable further steps in the testing-specific workflows. A standardized set of tags for scenarios, as defined in OpenLABEL for example, allows a test to specify scenarios based on a specific label. An ontology like OpenXOntology allows for further automation of such a workflow, allowing for scenario selection based on similar or related tags. This has similar implications for the interaction between an ODD and a scenario database, where tags specified in the ODD can be used to provide the search parameters for matching scenarios.

The OpenLABEL standard with its standardized format for labels for objects of interest (for sensors) is extremely beneficial to testing workflows that leverage different test techniques. Recorded data from a test drive on an open road for example can be used to automate the specification of scenarios for scenario-based testing (see Section 10.2 of the ASAM OpenXOntology standard for more details on this use case).

It can be expected that a standardized set of labels generally eases coverage determination, allowing for the implementation of automata that speed up the process.

Study group summary of this standard
As initial standards and guidelines on terminology for ADAS/AD are published it is expected that significant alignment efforts will be needed to realize the goals and advantages described above. These efforts will need to take place in and amongst the standardization organizations to prevent adding to the confusion in the industry. It is recommended that a coordinated effort to align the different taxonomies is initiated – initially at ASAM internally and in second step across organizations.