Special Issue on Safety of AI-Based Systems
Artificial Intelligence (AI) is applied more often in connected and automated (road) vehicles. For higher levels of automation, the application of AI-technology is indispensable. The full potential and limits of AI in this field are not fully understood yet. There is an imminent need for AI to be explainable, trustworthy, and responsible to enhance user acceptance and safety when deployed on public roads. In order to trust the decisions made by Connected Cooperative Automated Mobility (CCAM) using AI-technology, a deeper analysis of the control architecture design is needed, based on different performance indicators and beyond the existing testing framework for validation. Furthermore, methods for assessment and validation, i.e. the operational safety of AI-based systems, are in the early stages of development. It is of vital importance to address the safety assessment of AI robustness. This includes how the AI responds to a situation that it has not experienced yet and how well the AI behaviour, specifically the AI algorithm, aligns with the required intention of its Operational Design Domain (ODD).
To address these challenges, the journal seeks contributions on the following topics of interest, which include but are not limited to:
- The design and development of explainable and trustworthy AI-systems, the robustness of AI systems for CCAM applications. This can be on component level and on system level.
- Methods to ensure safety of AI, including e.g. the verification and validation of AI-based CCAM systems.
- The role of AI tools in data handling for development, validation and certification of CCAM applications
- AI and cybersecurity solutions for in-vehicle technologies (related to environmental perception and on-board decision making)
- Verification of AI for situational awareness
- Out-of-distribution detection, edge case detection
- AI self-awareness: the ability of an AI system to determine whether or not all its components are functioning according to specification
- Run-time verification and diagnostics of AI components
- Lifecycle management:
- how to handle functional updates of the system and adaptation of the environment or ethical, legal, and societal aspects
- how to evaluate self-adaptation of AI components
- Situation aware AI: the ability of an AI system to determine whether it can make trustworthy decisions in the current situation
- Value-aligned decision making under uncertainty for automated driving
For more information, please contact the Guest Editors:
Margriet VAN SCHIJNDEL–DE NOOIJ (Lead Guest Editor)
Eindhoven University of Technology, Netherlands
IFP Energies Nouvelles, France
TNO The Netherlands Institute for Applied Physics, The Netherlands
Olaf op den Camp
TNO The Netherlands Organization for Applied Scientific Research, The Netherlands
Graz University of Technology and Virtual Vehicle Research Center, Austria
Deadline to submit manuscripts for consideration:
October 31, 2023
Please submit your article at https://www.editorialmanager.com/saeconnautomveh and include a submission note in Editorial Manager to indicate that it is for this special issue.