Causal Models to Support Scenario-Based Testing of ADAS

Maier, Robert; Grabinger, Lisa; Urlhart, David; Mottok, Jürgen · 2023 · OpenAlex-citations

DOI: 10.1109/tits.2023.3317475

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 the reliability and safety of Advanced Driver Assistance Systems (ADAS) through scenario-based testing. As vehicle complexity increases, traditional testing methods struggle to account for the intricate interactions between vehicle components and environmental factors within an Operational Design Domain (ODD). The authors propose using Structural Causal Models (SCMs) as a framework to combine diverse knowledge sources—such as domain expertise, physical considerations, and observational data—into a machine-processable, probabilistic model. This approach aims to support the specification of test parameters and improve industrial system development processes by providing a mathematically sound basis for reasoning about system behavior. The study employs a simplified Advanced Emergency Braking System (AEBS) as a running example to demonstrate the practical application of causal models. The authors define key test parameters, including ego and agent vehicle velocities, distance, and Time To Collision (TTC). They construct a causal model that integrates these parameters with ODD factors, utilizing Bayesian Networks and SCMs to represent causal mechanisms and probability distributions. In the absence of sufficient observational data, the model is parametrized using literature, domain knowledge, and artificially generated data. The validity of the model’s inference results is then assessed using the CARLA virtual simulation platform, allowing for the translation of technical questions into probabilistic queries. The findings demonstrate that causal models can effectively guide the specification of test parameters for scenario-based testing. The model allows for the identification of relevant parameter instantiations and provides a causal rationale for prioritizing test configurations, particularly for "corner cases" or critical situations. By combining expert knowledge and data, the approach enables the derivation of test inputs that are traceable and certifiable, addressing the limitations of purely data-driven or expert-based methods. The simulation results validate that the model predictions are realistic and capable of capturing the complex interactions within the AEBS. The significance of this work lies in its contribution to the field of automotive safety and model-based engineering. It provides a concrete methodology for integrating causal inference into the V-model development lifecycle, bridging the gap between abstract safety requirements and concrete test scenarios. The paper highlights the potential of causal models to enhance the efficiency and comprehensiveness of safety testing by enabling the systematic identification of critical scenarios and supporting the validation of machine learning-based components. The authors conclude by discussing ongoing challenges, such as scalability and the integration of this approach into industrial workflows, and outline directions for future research to further refine the application of causal models in ADAS testing.

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: computational model