KI Absicherung produces White Paper on the state of the use of safe AI in autonomous driving
In subproject 3, a collaboration of 16 partners from the project consortium has produced an up-to-date overview of the use of trustworthy AI in autonomous driving. As one of the first papers of its kind, it captures the state-of-the-art of methods used to measure or mitigate safety concerns in modern machine learning methods.
The review paper presents these methods, evaluates them and identifies the current challenges and research directions in each case. Due to the focus of the partners involved, academic and less practical methods are described and examined by the experts in addition to methods and examples from the application domain.
The paper "Inspect, Understand, Overcome: A Survey of Practical Methods for AI Safety" is available for download here. Please direct any inquiries about it to Sebastian Houben.