Ontology based Scene Creation for the Development of Automated Vehicles

Bagschik, Gerrit; Menzel, Till; Maurer, Markus · 2018 · Crossref

DOI: 10.1109/ivs.2018.8500632

archive: archived pipeline: cataloged verified

Get this paper ↗ (DOI — opens at the source; we link to it, we don't host it)

Summary

This paper addresses the challenge of ensuring functional safety for automated vehicles, particularly those operating without permanent human supervision (Levels 3 and 4). Current safety assessments rely on expert-generated scenario catalogs, which are often incomplete and lack systematic traceability. The authors argue that to comply with standards like ISO 26262, developers must identify a comprehensive range of operating scenarios. To solve this, the paper proposes an ontology-based approach for systematically generating traffic scenes, moving beyond creative expert intuition to a knowledge-based, computer-aided process that ensures completeness and semantic consistency. The methodology utilizes ontologies, specifically the Web Ontology Language (OWL), to model traffic knowledge. The authors distinguish between terminological boxes (background knowledge, such as concepts and rules) and assertional boxes (situational knowledge, such as specific instances). They propose a five-layered model for scene representation: Layer 1 covers road geometry and topology; Layer 2 includes traffic infrastructure like signs and barriers; Layer 3 handles temporary manipulations like construction sites; Layer 4 defines dynamic objects and their maneuvers (e.g., lane changes, following); and Layer 5 accounts for environmental conditions like weather. Knowledge is acquired from German traffic guidelines, expert knowledge, and existing scenario catalogs. The system uses semantic web rule language to define behavioral rules and parameter relations, such as how weather affects road friction. The core finding is a process for deriving all possible valid assertional boxes (scenes) from the modeled terminological knowledge. Unlike previous works that use ontologies for real-time scene understanding from sensor data, this approach generates scenes *a priori* for simulation and safety analysis. The system combines entities from the five layers based on semantic constraints and dependencies defined in the ontology. For example, it ensures that traffic rules are correctly instantiated through infrastructure and that environmental effects logically influence physical parameters. The authors demonstrate that this method avoids the generation of physically impossible or unreasonable scenes, a common issue with database-based approaches that lack semantic understanding. The significance of this work lies in providing a scalable, traceable framework for scenario-based design in automated driving. By explicitly representing knowledge and rules, the approach supports hazard analysis and risk assessment by ensuring that all relevant scenarios are considered. It bridges the semantic gap in safety validation by providing a unified vocabulary and a systematic way to generate test cases for simulation. This allows developers to argue for the completeness of their safety analysis more rigorously than with expert-only methods, facilitating the transition from prototype development to series production of automated vehicles.

Provenance

The full processing record for this entry. Every stage of this paper's journey through the pipeline is logged — what ran, with which tool and model, how many attempts it took, and when it last completed.

Summary generated by qwen3.6-27b-prismaquant on 2026-06-26; verification: verified.

Topics

Ranked by relevance to this paper. Hover a topic for its definition.

• situational awareness

Information type

What kind of knowledge this paper contributes, grouped by family — independent of topic (what it is about) and method (how it was studied).

• Theoretical Contribution: conceptual framework