Grasping Causality for the Explanation of Criticality for Automated Driving

Abstract

Safeguarding automated driving systems at SAE levels 4 and 5 is a multi faceted challenge, for which classical distance-based approaches become infeasible. To alleviate this, contemporary scenario-based approaches suggest a decomposition into scenario classes combined with the statistical analysis of these classes regarding their criticality. Unfortunately, relying solely on associative statistics may fail to recognize the causalities leading to critical scenarios. These scenarios are prerequisite for the scenario-based development of safe automated driving systems. As to incorporate causal knowledge within the development process, this work introduces a formalization of causal queries. Answering these facilitates a causal understanding of safety-relevant influencing factors. This formalized causal knowledge can be used to specify and implement safety principles that provably reduce their associated criticality. Based on Judea Pearl’s causal theory, we define a causal relation as a causal structure together with a context, both related to a suitable domain ontology. The focus lies on modeling the effect of such influencing factors on criticality as measured by appropriate criticality metrics. Our main example is a causal relation for the influencing factor ‘reduced coefficient of friction’ and its effect on the Brake-Threat-Number. As availability and quality of data are important to answer the causal queries, we also discuss requirements on real-world and synthetic data acquisition. Overall, this work contributes to establish formal causal considerations within the safety process for automated driving systems.