Recent advancements in automated and autonomous driving systems have facilitated their integration into modern vehicles, enabling them to accurately perceive their surroundings and support or even fully undertake complex driving tasks. Given the complexity and unpredictable nature of driving environments and traffic situations, ensuring the correct behavior of such systems is essential to prevent hazardous situations, increase user acceptance, and avoid human harm. However, the increased complexity of these systems and the extensive search space of possible scenarios introduce significant challenges to testing and real-time fault management. Hence, besides rigorous testing during the development phase, there is a need for additional validation and verification during operation. This paper proposes utilizing Answer Set Programming (ASP), a form of declarative programming, for continuous real-time monitoring, fault detection, and explanation to ensure the correct functioning of automated and autonomous driving systems. Our approach aims to enhance the reliability and safety of such systems by detecting violations and providing explanations that can support fault-adaptive control or mitigation strategies. We demonstrate the effectiveness of our methodology across diverse scenarios executed within a simulation environment, discuss the main challenges encountered, and outline future research directions.

Abstract In this paper, we outline an approach for automatically generating challenging road networks for virtual testing of an automated lane‐keeping system. Based on a set of control points, we construct a parametric curve representing a road network, defining the dynamic driving task an automated lane‐keeping system‐equipped vehicle must perform. Changing control points has a global influence on the resulting road geometry. Our approach uses search to find control‐point sets that result in a challenging road, eventually forcing the vehicle to leave the intended path. We apply our approach in different search variants to evaluate their performance regarding test efficiency and the diversity of failing tests. In addition, we evaluate different genetic algorithm control parameter configurations to investigate the most influential parameters and if specific configurations can be seen asoptimal, leading to better results than others. For both studies, we consider another search‐based test method and two different random test generators as a baseline for comparison. The empirical results indicate that specific control parameter settings increase the overall performance for each search variant. While the population size is the most influential control parameter for all methods, the performance improvement when usingoptimalsettings is only significant for one method.

Vision systems, i.e., systems that enable the detection and tracking of objects in images, have gained substantial importance over the past decades. They are used in quality assurance applications, e.g., for finding surface defects in products during manufacturing, surveillance, but also automated driving, requiring reliable behavior. Interestingly, there is only little work on quality assurance and especially testing of vision systems in general. In this paper, we contribute to the area of testing vision software, and present a framework for the automated generation of tests for systems based on vision and image recognition with the focus on easy usage, uniform usability and expandability. The framework makes use of existing libraries for modifying the original images and to obtain similarities between the original and modified images. We show how such a framework can be used for testing a particular industrial application on identifying defects on riblet surfaces and present preliminary results from the image classification domain.

Testing AI-based systems and especially when they rely on machine learning is considered a challenging task. In this paper, we contribute to this challenge considering testing neural networks utilizing mutation testing. A former paper focused on applying mutation testing to the configuration of neural networks leading to the conclusion that mutation testing can be effectively used. In this paper, we discuss a substantially extended empirical evaluation where we considered different test data and the source code of neural network implementations. In particular, we discuss whether a mutated neural network can be distinguished from the original one after learning, only considering a test evaluation. Unfortunately, this is rarely the case leading to a low mutation score. As a consequence, we see that the testing method, which works well at the configuration level of a neural network, is not sufficient to test neural network libraries requiring substantially more testing effort for assuring quality.