Safety Concept for Autonomous Vehicles

Reschka, Andreas · 2016 · Crossref

DOI: 10.1007/978-3-662-48847-8_23

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 critical gap in safety concepts for autonomous vehicles, noting that current development focuses primarily on vehicle guidance functionality rather than comprehensive safety architectures. While numerous experimental projects have demonstrated impressive capabilities, they rely on safety drivers to monitor systems and intervene during faults, effectively classifying them as semi-automated. The author argues that for future fully autonomous systems to operate independently without human monitoring, a robust safety concept covering specification, design, and functional testing is required. Central to this argument is the definition of a "safe state," which, according to ISO 26262, is a condition where risk is below a socially accepted threshold. The paper highlights the challenge of dynamically determining this threshold based on real-time situational factors, such as traffic objects, legal conditions, and vehicle capabilities. The study employs a comparative review of existing safety mechanisms across various domains to identify suitable approaches for autonomous driving. It analyzes safety concepts in series-production driver assistance systems, experimental autonomous vehicle projects (such as DARPA, Stadtpilot, and Google), and other disciplines including track vehicles, aviation (X-by-Wire systems), robotics, and nuclear power stations. The analysis reveals that current experimental vehicles largely rely on handing control to a safety driver or stopping immediately as their primary safety measures. In contrast, other fields offer more sophisticated strategies: track vehicles utilize infrastructure-centric coordination and redundant monitoring; aviation employs triple-redundant control circuits; and robotics use hierarchical safety policies to override complex actions. The paper suggests that autonomous vehicles may require similar levels of redundancy and standardized safety specifications, akin to nuclear power plants, to handle complex error chains. The findings indicate that no single comprehensive safety concept currently exists for driverless vehicles on public roads. The paper evaluates specific use cases to define appropriate safe states. For freeway driving with a safety driver, safe states include stopping on the shoulder or handing control to the driver. For autonomous valet parking without a driver, the lack of human intervention poses significant challenges, particularly regarding securing the vehicle if it stops in hazardous locations or blocks emergency access. The author concludes that simply stopping or relying on a safety driver is insufficient for full automation. Instead, autonomous vehicles must be capable of leaving traffic flow and stopping safely at the roadside independently. This requires reliable environment perception, decision-making, and potentially remote operator assistance or redundant hardware systems to ensure that the vehicle can attain a safe state under all conditions.

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.

• driverless ads
• situational awareness
• automation surprise
• mode 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