Context: The AI Act marks a new chapter in AI governance, affecting companies around the world seeking to offer their services within the European Union. This study focuses on the comprehensive AI Act requirements set out for high-risk AI systems. Objectives: We explored the perceived compliance challenge for the AI Act’s high-risk requirements and associated contributing factors; the AI Act’s impact on industry in terms of positive and negative side effects; and the sentiment of industry practitioners towards the AI Act’s Codes of Conduct for the voluntary application of the act’s high-risk AI requirements. Method: A multiple case study encompassing six case companies supplemented by three independent experts with a total of 16 respondents was conducted. Results: A ranking represents the different perceived levels of challenge for each AI Act high-risk requirement. The ranking is led by the following requirements, starting with the most challenging one: (1) data quality and governance (Art 10), (2) accuracy, robustness, and cybersecurity (Art 15), (3) risk and quality management system (Art 9, 17), and (4) transparency (Art 13). Moreover, four contributing factors emerged that impact the perceived compliance challenge: (1) industry and brand values, (2) existing regulatory environment, (3) AI maturity level and proficiency, and (4) company size. We identified several general key factors for the AI Act’s impact on industry and outlined strong arguments both for and against the AI Act voiced by practitioners. The sentiment towards the AI Act’s Codes of Conduct turned out very positive. Conclusion: This study offers a valuable primary research contribution to software engineering, where the state-of-the-art remains short of compliance-oriented studies with a focus on the operationalization of certain AI Act aspects. Future work is advised to develop artifacts facilitating AI Act operationalization and to validate them with industry partners.

Abstract With the emergence of autonomous vehicles comes the requirement of adequate and rigorous testing, particularly in critical scenarios that are both challenging and potentially hazardous. Generating synthetic simulation-based critical scenarios for testing autonomous vehicles has therefore received considerable interest, yet it is unclear how such scenarios relate to the actual crash or near-crash scenarios in the real world. Consequently, their realism is unknown. In this paper, we define realism as the degree of similarity of synthetic critical scenarios to real-world critical scenarios. We propose a methodology to measure realism using two metrics, namely attribute distribution and Euclidean distance. The methodology extracts various attributes from synthetic and realistic critical scenario datasets and performs a set of statistical tests to compare their distributions and distances. As a proof of concept for our methodology, we compare synthetic collision scenarios from DeepScenario against realistic autonomous vehicle collisions collected by the Department of Motor Vehicles in California, to analyse how well DeepScenario synthetic collision scenarios are aligned with real autonomous vehicle collisions recorded in California. We focus on five key attributes that are extractable from both datasets, and analyse the attribution distribution and distance between scenarios in the two datasets. Further, we derive recommendations to improve the realism of synthetic scenarios based on our analysis. Our study of realism provides a framework that can be replicated and extended for other dataset both concerning real-world and synthetically-generated scenarios.

Background: In recent years, a discourse on how to systematically consider and report threats to validity started to gain momentum within the empirical software engineering community. Aims: With this study, we aim to systematically underpin the current state of threats to validity practices in software engineering research. Method: We conduct a literature review comprising 91 papers awarded with the ACM SIGSOFT Distinguished Paper Award at the ACM/IEEE International Conference on Software Engineering. Data is extracted and analyzed by considering six main facets of threats to validity, e.g., their explicit documentation, categorization, discussion of limitations, and trade-offs. Results: Results corroborate current critiques to the threats management state of the art. Threats result to be seldom discussed in depth, and are mostly considered as an enforced afterthought rather than an active concern of the research design and execution. Conclusions: To improve the observed practice, we derived items to consider for researchers, reviewers and readers, and call for a community action to increase the understanding of knowledge creation in empirical software engineering research.

Testing autonomous driving systems for safety and reliability is essential, yet complex. A primary challenge is identifying relevant test scenarios, especially the critical ones that may expose hazards or harm to autonomous vehicles and other road users. Although numerous approaches and tools for critical scenario identification are proposed, the industry practices for selection, implementation, and evaluation of approaches, are not well understood. Therefore, we aim at exploring practical aspects of how autonomous driving systems are tested, particularly the identification and use of critical scenarios. We interviewed 13 practitioners from 7 companies in autonomous driving in Sweden. We used thematic modeling to analyse and synthesize the interview data. As a result, we present 9 themes of practices and 4 themes of challenges related to critical scenarios. Our analysis indicates there is little joint effort in the industry, despite every approach has its own limitations, and tools and platforms are lacking. To that end, we recommend the industry and academia combine different approaches, collaborate among different stakeholders, and continuously learn the field. The contributions of our study are exploration and synthesis of industry practices and related challenges for critical scenario identification and testing, and potential increase of industry relevance for future studies.

Effective industry-academia collaboration may increase software engineering research relevance by increased realism, yet very challenging for reasons like confidentiality concerns, different objectives and priorities. We analyse industry-academia collaboration scenarios based on our own experiences as Ph.D. student and supervisor, and provide insights and recommendations to facilitate future collaborations with industry. We first present our industry-academia collaboration experiences that span over two and a half years with different companies. Then, we analyse both facilitators and problems from those scenarios and synthesize recommendations based on that. Five different scenarios are analysed, including both success and failure scenarios. Reflections and insights into these experiences as well as some general recommendations are presented. We believe such experiences and insights are helpful for academic researchers to pursue industry-academia collaboration. We plan to continuously report our experience and provide our suggestions for effective collaboration with industry.

Abstract Autonomous driving has become an important research area for road traffic, whereas testing of autonomous driving systems to ensure a safe and reliable operation remains an open challenge. Substantial real-world testing or massive driving data collection does not scale since the potential test scenarios in real-world traffic are infinite, and covering large shares of them in the test is impractical. Thus, critical ones have to be prioritized. We have developed an approach for critical test scenario identification and in this study, we implement the approach and validate it on two real autonomous driving systems from industry by integrating it into their tool-chain. Our main contribution in this work is the demonstration and validation of our approach for critical scenario identification for testing real autonomous driving systems.

There is a growing interest in industry and academia in machine learning (ML) testing. We believe that industry and academia need to learn together to produce rigorous and relevant knowledge. In this study, we initiate a collaboration between stakeholders from one case company, one research institute, and one university. To establish a common view of the problem domain, we applied an interactive rapid review of the state of the art. Four researchers from Lund University and RISE Research Institutes and four practitioners from Axis Communications reviewed a set of 180 primary studies on ML testing. We developed a taxonomy for the communication around ML testing challenges and results and identified a list of 12 review questions relevant for Axis Communications. The three most important questions (data testing, metrics for assessment, and test generation) were mapped to the literature, and an in-depth analysis of the 35 primary studies matching the most important question (data testing) was made. A final set of the five best matches were analysed and we reflect on the criteria for applicability and relevance for the industry. The taxonomies are helpful for communication but not final. Furthermore, there was no perfect match to the case company's investigated review question (data testing). However, we extracted relevant approaches from the five studies on a conceptual level to support later context-specific improvements. We found the interactive rapid review approach useful for triggering and aligning communication between the different stakeholders.

While autonomous driving systems are expected to change future means of mobility and reduce road accidents, understanding intensive and complex traffic situations is essential to enable testing of such systems under realistic traffic conditions. Particularly, we need to cover more relevant driving scenarios in the test. However, we do not want to spend time and resources testing useless scenarios that never happen in the real road traffic. In this work, we propose a new model that defines the distribution of scenarios using TTC (Time-to-Collision) for the vehicle–pedestrian interactions at unsignalized crossings based on the traffic density. The scenario distribution can be used as an input for test scenario generation and selection. We validate the model using real traffic data collected in Sweden and the result indicates that the model is effective and consistently upholds the real distribution, especially for critical scenarios with TTC less than 3 seconds. We also demonstrate the use of the model by connecting it to the testing of an auto-braking function from the industry. As a first step, our contribution is a model that predicts the worst-case distribution of scenarios using TTC and provides a mandatory input for testing autonomous driving systems.