A Concept to Support AI Models by Using Ontologies - Presented on the Basis of German Technical Specifications for Lane Markings

Artificial Intelligence (AI) and Machine Learning (ML) deliver promising approaches to the development of assisted as well as automated and autonomous driving technologies. However, learning all possible traffic situations and outcomes is almost not feasible. Furthermore, machine learning-based models are usually regarded as a black box and one cannot trace their decisions for a certain behavior. To counteract this, the authors propose an ontology-based model, which integrates normative knowledge, to support the decision making of the AI for automated and autonomous vehicles. Since traffic rules and laws are explicitly defined in the model, the authors can easily track any derived decisions, eliminating the necessity of learning all possible traffic situations. The authors formalize the German Technical Specifications on Lane Markings into an ontology for a better representation of the traffic environment and thus improve the situational awareness of automated and autonomous vehicles. Additionally, the reasoning capacity of an ontology based-model allows for deriving concepts in multiple ways, which can serve as redundant information about lane and lane markings to enhance the understanding of the traffic situation. Finally, in contrast to learning-based models, our transparent ontology-based model allows for the validation and verification of automated and autonomous systems and vehicles.

Language

  • English

Media Info

  • Media Type: Web
  • Features: Figures; References; Tables;
  • Pagination: 14p
  • Monograph Title: 27th International Technical Conference on the Enhanced Safety of Vehicles (ESV): Enhanced and Equitable Vehicle Safety for All: Toward the Next 50 Years

Subject/Index Terms

Filing Info

  • Accession Number: 01892755
  • Record Type: Publication
  • Report/Paper Numbers: 23-0265
  • Files: TRIS, ATRI, USDOT
  • Created Date: Sep 11 2023 11:42AM