Beatriz Cabrero-Daniel

Kıvanç Tatar, Petter Ericson, Kelsey Cotton et al.

The explosion of content generated by artificial intelligence (AI) models has initiated a cultural shift in arts, music, and media, whereby roles are changing, values are shifting, and conventions are challenged. The vast, readily available dataset of the Internet has created an environment for AI models to be trained on any content on the Web. With AI models shared openly and used by many globally, how does this new paradigm shift challenge the status quo in artistic practices? What kind of changes will AI technology bring to music, arts, and new media?

Malsha Ashani Mahawatta Dona, Beatriz Cabrero-Daniel, Yinan Yu et al.

Today's Large Language Models (LLMs) have showcased exemplary capabilities, ranging from simple text generation to advanced image processing. Such models are currently being explored for in-vehicle services such as supporting perception tasks in Advanced Driver Assistance Systems (ADAS) or Autonomous Driving (AD) systems, given the LLMs' capabilities to process multi-modal data. However, LLMs often generate nonsensical or unfaithful information, known as ``hallucinations'': a notable issue that needs to be mitigated. In this paper, we systematically explore the adoption of SelfCheckGPT to spot hallucinations by three state-of-the-art LLMs (GPT-4o, LLaVA, and Llama3) when analysing visual automotive data from two sources: Waymo Open Dataset, from the US, and PREPER CITY dataset, from Sweden. Our results show that GPT-4o is better at generating faithful image captions than LLaVA, whereas the former demonstrated leniency in mislabeling non-hallucinated content as hallucinations compared to the latter. Furthermore, the analysis of the performance metrics revealed that the dataset type (Waymo or PREPER CITY) did not significantly affect the quality of the captions or the effectiveness of hallucination detection. However, the models showed better performance rates over images captured during daytime, compared to during dawn, dusk or night. Overall, the results show that SelfCheckGPT and its adaptation can be used to filter hallucinations in generated traffic-related image captions for state-of-the-art LLMs.

Malsha Ashani Mahawatta Dona, Beatriz Cabrero-Daniel, Yinan Yu et al.

Today's advanced driver assistance systems (ADAS), like adaptive cruise control or rear collision warning, are finding broader adoption across vehicle classes. Integrating such advanced, multimodal Large Language Models (LLMs) on board a vehicle, which are capable of processing text, images, audio, and other data types, may have the potential to greatly enhance passenger comfort. Yet, an LLM's hallucinations are still a major challenge to be addressed. In this paper, we systematically assessed potential hallucination detection strategies for such LLMs in the context of object detection in vision-based data on the example of pedestrian detection and localization. We evaluate three hallucination detection strategies applied to two state-of-the-art LLMs, the proprietary GPT-4V and the open LLaVA, on two datasets (Waymo/US and PREPER CITY/Sweden). Our results show that these LLMs can describe a traffic situation to an impressive level of detail but are still challenged for further analysis activities such as object localization. We evaluate and extend hallucination detection approaches when applying these LLMs to video sequences in the example of pedestrian detection. Our experiments show that, at the moment, the state-of-the-art proprietary LLM performs much better than the open LLM. Furthermore, consistency enhancement techniques based on voting, such as the Best-of-Three (BO3) method, do not effectively reduce hallucinations in LLMs that tend to exhibit high false negatives in detecting pedestrians. However, extending the hallucination detection by including information from the past helps to improve results.

Malsha Ashani Mahawatta Dona, Beatriz Cabrero-Daniel, Yinan Yu et al.

Today's advanced automotive systems are turning into intelligent Cyber-Physical Systems (CPS), bringing computational intelligence to their cyber-physical context. Such systems power advanced driver assistance systems (ADAS) that observe a vehicle's surroundings for their functionality. However, such ADAS have clear limitations in scenarios when the direct line-of-sight to surrounding objects is occluded, like in urban areas. Imagine now automated driving (AD) systems that ideally could benefit from other vehicles' field-of-view in such occluded situations to increase traffic safety if, for example, locations about pedestrians can be shared across vehicles. Current literature suggests vehicle-to-infrastructure (V2I) via roadside units (RSUs) or vehicle-to-vehicle (V2V) communication to address such issues that stream sensor or object data between vehicles. When considering the ongoing revolution in vehicle system architectures towards powerful, centralized processing units with hardware accelerators, foreseeing the onboard presence of large language models (LLMs) to improve the passengers' comfort when using voice assistants becomes a reality. We are suggesting and evaluating a concept to complement the ego vehicle's field-of-view (FOV) with another vehicle's FOV by tapping into their onboard LLM to let the machines have a dialogue about what the other vehicle ``sees''. Our results show that very recent versions of LLMs, such as GPT-4V and GPT-4o, understand a traffic situation to an impressive level of detail, and hence, they can be used even to spot traffic participants. However, better prompts are needed to improve the detection quality and future work is needed towards a standardised message interchange format between vehicles.

Junyi Gu, Beatriz Cabrero-Daniel, Ali Nouri et al.

We have recently observed the commercial roll-out of robotaxis in various countries. They are deployed within an operational design domain (ODD) on specific routes and environmental conditions, and are subject to continuous monitoring to regain control in safety-critical situations. Since ODDs typically cover urban areas, robotaxis must reliably detect vulnerable road users (VRUs) such as pedestrians, bicyclists, or e-scooter riders. To better handle such varied traffic situations, end-to-end AI, which directly compute vehicle control actions from multi-modal sensor data instead of only for perception, is on the rise. High quality data is needed for systematically training and evaluating such systems within their OOD. In this work, we propose PCICF, a framework to systematically identify and classify VRU situations to support ODD's incident analysis. We base our work on the existing synthetic dataset SMIRK, and enhance it by extending its single-pedestrian-only design into the MoreSMIRK dataset, a structured dictionary of multi-pedestrian crossing situations constructed systematically. We then use space-filling curves (SFCs) to transform multi-dimensional features of scenarios into characteristic patterns, which we match with corresponding entries in MoreSMIRK. We evaluate PCICF with the large real-world dataset PIE, which contains more than 150 manually annotated pedestrian crossing videos. We show that PCICF can successfully identify and classify complex pedestrian crossings, even when groups of pedestrians merge or split. By leveraging computationally efficient components like SFCs, PCICF has even potential to be used onboard of robotaxis for OOD detection for example. We share an open-source replication package for PCICF containing its algorithms, the complete MoreSMIRK dataset and dictionary, as well as our experiment results presented in: https://github.com/Claud1234/PCICF

Ali Nouri, Beatriz Cabrero-Daniel, Fredrik Törner et al.

Changes and updates in the requirement artifacts, which can be frequent in the automotive domain, are a challenge for SafetyOps. Large Language Models (LLMs), with their impressive natural language understanding and generating capabilities, can play a key role in automatically refining and decomposing requirements after each update. In this study, we propose a prototype of a pipeline of prompts and LLMs that receives an item definition and outputs solutions in the form of safety requirements. This pipeline also performs a review of the requirement dataset and identifies redundant or contradictory requirements. We first identified the necessary characteristics for performing HARA and then defined tests to assess an LLM's capability in meeting these criteria. We used design science with multiple iterations and let experts from different companies evaluate each cycle quantitatively and qualitatively. Finally, the prototype was implemented at a case company and the responsible team evaluated its efficiency.

Krishna Ronanki, Beatriz Cabrero-Daniel, Tomas Herda et al.

Context: Large language models (LLMs) are observed to have a significant positive impact on various software engineering (SE) activities. With improved accessibility, the adoption of powerful LLMs in industry has surged recently. However, there is a lack of actionable best practices for the efficient and responsible adoption of LLMs within industrial software settings. Objectives: We developed seven actionable recommendations to address this research gap. Methods: We conducted a multi-case study with three organisations that use LLMs within their SE activities and synthesised seven recommendations through qualitative thematic analysis. We conducted a complementary online survey with software practitioners from various industries to evaluate the perceived relevance of our recommendations. Results: Our results and recommendations focus on (i) users' preference to use LLMs as AI assistants, (ii) the importance of relevant stakeholders' satisfaction in the LLM-output evaluation, (iii) scoping the applicability of LLMs within SE tasks, (iv) the effect of LLMs on SE workflows, (v) the necessity and directions for developing human oversight mechanisms, and (vi) the necessary skills for practitioners for leveraging LLMs within SE. The online survey indicates a high level of agreement from the participants regarding the perceived relevance of the recommendations. Conclusion: We outline future research directions, including mapping the seven recommendations to the principles of the EU AI Act (AIA) in order to examine how they relate to the current regulatory compliance frameworks.

Beatriz Cabrero-Daniel, Tomas Herda, Victoria Pichler et al.

This action research study focuses on the integration of "AI assistants" in two Agile software development meetings: the Daily Scrum and a feature refinement, a planning meeting that is part of an in-house Scaled Agile framework. We discuss the critical drivers of success, and establish a link between the use of AI and team collaboration dynamics. We conclude with a list of lessons learnt during the interventions in an industrial context, and provide a assessment checklist for companies and teams to reflect on their readiness level. This paper is thus a road-map to facilitate the integration of AI tools in Agile setups.

Krishna Ronanki, Beatriz Cabrero-Daniel, Christian Berger

Recent Generative Artificial Intelligence (GenAI) trends focus on various applications, including creating stories, illustrations, poems, articles, computer code, music compositions, and videos. Extrinsic hallucinations are a critical limitation of such GenAI, which can lead to significant challenges in achieving and maintaining the trustworthiness of GenAI. In this paper, we propose two new concepts that we believe will aid the research community in addressing limitations associated with the application of GenAI models. First, we propose a definition for the "desirability" of GenAI outputs and three factors which are observed to influence it. Second, drawing inspiration from Martin Fowler's code smells, we propose the concept of "prompt smells" and the adverse effects they are observed to have on the desirability of GenAI outputs. We expect our work will contribute to the ongoing conversation about the desirability of GenAI outputs and help advance the field in a meaningful way.

Ali Nouri, Johan Andersson, Kailash De Jesus Hornig et al.

Automated Driving System (ADS) is a safety-critical software system responsible for the interpretation of the vehicle's environment and making decisions accordingly. The unbounded complexity of the driving context, including unforeseeable events, necessitate continuous improvement, often achieved through iterative DevOps processes. However, DevOps processes are themselves complex, making these improvements both time- and resource-intensive. Automation in code generation for ADS using Large Language Models (LLM) is one potential approach to address this challenge. Nevertheless, the development of ADS requires rigorous processes to verify, validate, assess, and qualify the code before it can be deployed in the vehicle and used. In this study, we developed and evaluated a prototype for automatic code generation and assessment using a designed pipeline of a LLM-based agent, simulation model, and rule-based feedback generator in an industrial setup. The LLM-generated code is evaluated automatically in a simulation model against multiple critical traffic scenarios, and an assessment report is provided as feedback to the LLM for modification or bug fixing. We report about the experimental results of the prototype employing Codellama:34b, DeepSeek (r1:32b and Coder:33b), CodeGemma:7b, Mistral:7b, and GPT4 for Adaptive Cruise Control (ACC) and Unsupervised Collision Avoidance by Evasive Manoeuvre (CAEM). We finally assessed the tool with 11 experts at two Original Equipment Manufacturers (OEMs) by conducting an interview study.

Malsha Ashani Mahawatta Dona, Beatriz Cabrero-Daniel, Yinan Yu et al.

Large language models (LLMs) are growingly extended to process multimodal data such as text and video simultaneously. Their remarkable performance in understanding what is shown in images is surpassing specialized neural networks (NNs) such as Yolo that is supporting only a well-formed but very limited vocabulary, ie., objects that they are able to detect. When being non-restricted, LLMs and in particular state-of-the-art vision language models (VLMs) show impressive performance to describe even complex traffic situations. This is making them potentially suitable components for automotive perception systems to support the understanding of complex traffic situations or edge case situation. However, LLMs and VLMs are prone to hallucination, which mean to either potentially not seeing traffic agents such as vulnerable road users who are present in a situation, or to seeing traffic agents who are not there in reality. While the latter is unwanted making an ADAS or autonomous driving systems (ADS) to unnecessarily slow down, the former could lead to disastrous decisions from an ADS. In our work, we are systematically assessing the performance of 3 state-of-the-art VLMs on a diverse subset of traffic situations sampled from the Waymo Open Dataset to support safety guardrails for capturing such hallucinations in VLM-supported perception systems. We observe that both, proprietary and open VLMs exhibit remarkable image understanding capabilities even paying thorough attention to fine details sometimes difficult to spot for us humans. However, they are also still prone to making up elements in their descriptions to date requiring hallucination detection strategies such as BetterCheck that we propose in our work.

Ali Nouri, Beatriz Cabrero-Daniel, Zhennan Fei et al.

Software engineers in various industrial domains are already using Large Language Models (LLMs) to accelerate the process of implementing parts of software systems. When considering its potential use for ADAS or AD systems in the automotive context, there is a need to systematically assess this new setup: LLMs entail a well-documented set of risks for safety-related systems' development due to their stochastic nature. To reduce the effort for code reviewers to evaluate LLM-generated code, we propose an evaluation pipeline to conduct sanity-checks on the generated code. We compare the performance of six state-of-the-art LLMs (CodeLlama, CodeGemma, DeepSeek-r1, DeepSeek-Coders, Mistral, and GPT-4) on four safety-related programming tasks. Additionally, we qualitatively analyse the most frequent faults generated by these LLMs, creating a failure-mode catalogue to support human reviewers. Finally, the limitations and capabilities of LLMs in code generation, and the use of the proposed pipeline in the existing process, are discussed.

Ali Nouri, Beatriz Cabrero-Daniel, Fredrik Törner et al.

DevOps is a necessity in many industries, including the development of Autonomous Vehicles. In those settings, there are iterative activities that reduce the speed of SafetyOps cycles. One of these activities is "Hazard Analysis & Risk Assessment" (HARA), which is an essential step to start the safety requirements specification. As a potential approach to increase the speed of this step in SafetyOps, we have delved into the capabilities of Large Language Models (LLMs). Our objective is to systematically assess their potential for application in the field of safety engineering. To that end, we propose a framework to support a higher degree of automation of HARA with LLMs. Despite our endeavors to automate as much of the process as possible, expert review remains crucial to ensure the validity and correctness of the analysis results, with necessary modifications made accordingly.

Beatriz Cabrero-Daniel, Andrea Sanagustín Cabrero

Natural Language Generation tools, such as chatbots that can generate human-like conversational text, are becoming more common both for personal and professional use. However, there are concerns about their trustworthiness and ethical implications. The paper addresses the problem of understanding how different users (e.g., linguists, engineers) perceive and adopt these tools and their perception of machine-generated text quality. It also discusses the perceived advantages and limitations of Natural Language Generation tools, as well as users' beliefs on governance strategies. The main findings of this study include the impact of users' field and level of expertise on the perceived trust and adoption of Natural Language Generation tools, the users' assessment of the accuracy, fluency, and potential biases of machine-generated text in comparison to human-written text, and an analysis of the advantages and ethical risks associated with these tools as identified by the participants. Moreover, this paper discusses the potential implications of these findings for enhancing the AI development process. The paper sheds light on how different user characteristics shape their beliefs on the quality and overall trustworthiness of machine-generated text. Furthermore, it examines the benefits and risks of these tools from the perspectives of different users.

Krishna Ronanki, Beatriz Cabrero-Daniel, Jennifer Horkoff et al.

Complying with the EU AI Act (AIA) guidelines while developing and implementing AI systems will soon be mandatory within the EU. However, practitioners lack actionable instructions to operationalise ethics during AI systems development. A literature review of different ethical guidelines revealed inconsistencies in the principles addressed and the terminology used to describe them. Furthermore, requirements engineering (RE), which is identified to foster trustworthiness in the AI development process from the early stages was observed to be absent in a lot of frameworks that support the development of ethical and trustworthy AI. This incongruous phrasing combined with a lack of concrete development practices makes trustworthy AI development harder. To address this concern, we formulated a comparison table for the terminology used and the coverage of the ethical AI principles in major ethical AI guidelines. We then examined the applicability of ethical AI development frameworks for performing effective RE during the development of trustworthy AI systems. A tertiary review and meta-analysis of literature discussing ethical AI frameworks revealed their limitations when developing trustworthy AI. Based on our findings, we propose recommendations to address such limitations during the development of trustworthy AI.

Beatriz Cabrero-Daniel

This study explores the benefits and challenges of integrating Artificial Intelligence with Agile software development methodologies, focusing on improving continuous integration and delivery. A systematic literature review and longitudinal meta-analysis of the retrieved studies was conducted to analyse the role of Artificial Intelligence and it's future applications within Agile software development. The review helped identify critical challenges, such as the need for specialised socio-technical expertise. While Artificial Intelligence holds promise for improved software development practices, further research is needed to better understand its impact on processes and practitioners, and to address the indirect challenges associated with its implementation.