Jan Bosch

Hongyi Zhang, Jan Bosch, Helena Holmström Olsson

Federated Learning (FL) has emerged as a promising approach for collaborative machine learning, addressing data privacy concerns. However, existing FL platforms and frameworks often present challenges for software engineers in terms of complexity, limited customization options, and scalability limitations. In this paper, we introduce EdgeFL, an edge-only lightweight decentralized FL framework, designed to overcome the limitations of centralized aggregation and scalability in FL deployments. By adopting an edge-only model training and aggregation approach, EdgeFL eliminates the need for a central server, enabling seamless scalability across diverse use cases. With a straightforward integration process requiring just four lines of code (LOC), software engineers can easily incorporate FL functionalities into their AI products. Furthermore, EdgeFL offers the flexibility to customize aggregation functions, empowering engineers to adapt them to specific needs. Based on the results, we demonstrate that EdgeFL achieves superior performance compared to existing FL platforms/frameworks. Our results show that EdgeFL reduces weights update latency and enables faster model evolution, enhancing the efficiency of edge devices. Moreover, EdgeFL exhibits improved classification accuracy compared to traditional centralized FL approaches. By leveraging EdgeFL, software engineers can harness the benefits of federated learning while overcoming the challenges associated with existing FL platforms/frameworks.

Choro Ulan uulu, Mikhail Kulyabin, Iris Fuhrmann et al.

Critical domain knowledge typically resides with few experts, creating organizational bottlenecks in scalability and decision-making. Non-experts struggle to create effective visualizations, leading to suboptimal insights and diverting expert time. This paper investigates how to capture and embed human domain knowledge into AI agent systems through an industrial case study. We propose a software engineering framework to capture human domain knowledge for engineering AI agents in simulation data visualization by augmenting a Large Language Model (LLM) with a request classifier, Retrieval-Augmented Generation (RAG) system for code generation, codified expert rules, and visualization design principles unified in an agent demonstrating autonomous, reactive, proactive, and social behavior. Evaluation across five scenarios spanning multiple engineering domains with 12 evaluators demonstrates 206% improvement in output quality, with our agent achieving expert-level ratings in all cases versus baseline's poor performance, while maintaining superior code quality with lower variance. Our contributions are: an automated agent-based system for visualization generation and a validated framework for systematically capturing human domain knowledge and codifying tacit expert knowledge into AI agents, demonstrating that non-experts can achieve expert-level outcomes in specialized domains.

Spyridon Alvanakis Apostolou, Jan Bosch, Helena Holmström Olsson

Agentic AI in software product development is increasingly adopted by organizations, yet the field lacks a consolidated synthesis of where adoption is mature, which architectural patterns dominate, and what limitations and coping mechanisms exist in industrial deployments. This systematic literature review addresses these gaps by establishing a body of knowledge as a starting point. Following Kitchenham guidelines, we queried four major research databases, obtaining over 1600 candidate publications. To handle this volume, we developed and validated a domain-agnostic multi-agent screening pipeline that extends prior LLM-assisted review tools by combining automatic metadata curation, inter-agent iterative dialogue, and conflict-resolution defaults that minimize false negatives. From the 92 manually verified primary studies, our thematic synthesis reveals that output verifiability is the primary enabler of agentic adoption: later SDLC phases, whose outputs are objectively evaluable through executable feedback, demonstrate the highest maturity and industrial presence, while earlier phases remain almost exclusively academic proofs-of-concept. We identify the Planner-Executor-Reviewer role specialization as the dominant architectural pattern, with the Reviewer agent implementing verifiability through executable feedback loops. Across all challenge categories, industrial mitigation strategies converge on confining agent actions to verifiable, bounded spaces. This study contributes a comprehensive characterization of the current literature on agentic systems in software product development, and a methodological contribution in the form of an AI-assisted tool to automate the screening phase in high-volume SLR domains.

Spyridon Alvanakis Apostolou, Jan Bosch, Helena Holmström Olsson

Agentic AI systems are entering software engineering workflows, yet empirical evidence on how industrial organizations actually adopt them remains sparse. We present a qualitative interview study with sixteen practitioners across twelve companies of varying size and domain. This study characterizes the current agentic AI adoption state of these companies, employing a six-level maturity framework adapted from established AI-driven organizations. The findings reveal that seven companies operate at Level~1 (AI Assistants), four companies at Level~2 (AI Compensators), and only one in Level~3 (Multi-Agent Orchestration), with large and safety-regulated organizations among the most advanced adopters. The primary finding is a capability-deployment verification gap, four companies demonstrated higher-level experimental AI capabilities but cannot integrate them into production workflows because adequate output verification mechanisms are absent, leaving human-in-the-loop as the only trusted verification mechanism. This gap is shaped by four recurring barriers: context window of LLMs constraints especially when diverse knowledge aggregation is needed, under-performance on proprietary programming languages and protocols, non-determinism incompatible with qualification standards, and data confidentiality concerns. Two interdependent dimensions of this gap emerge from these findings (information asymmetry and qualification absence) framing a core open problem for industrial agentic integration.

Rimma Dzhusupova, Jan Bosch, Helena Holmstrom Olsson

The Engineering, Procurement and Construction (EPC) businesses operating within the energy sector are recognizing the increasing importance of Artificial Intelligence (AI). Many EPC companies and their clients have realized the benefits of applying AI to their businesses in order to reduce manual work, drive productivity, and streamline future operations of engineered installations in a highly competitive industry. The current AI market offers various solutions and services to support this industry, but organizations must understand how to acquire AI technology in the most beneficial way based on their business strategy and available resources. This paper presents a framework for EPC companies in their transformation towards AI. Our work is based on examples of project execution of AI-based products development at one of the biggest EPC contractors worldwide and on insights from EPC vendor companies already integrating AI into their engineering solutions. The paper covers the entire life cycle of building AI solutions, from initial business understanding to deployment and further evolution. The framework identifies how various factors influence the choice of approach toward AI project development within large international engineering corporations. By presenting a practical guide for optimal approach selection, this paper contributes to the research in AI project management and organizational strategies for integrating AI technology into businesses. The framework might also help engineering companies choose the optimum AI approach to create business value.

Mikhail Kulyabin, Jan Joosten, Choro Ulan uulu et al.

Customer feedback in industrial forums reflect a rich but underexplored source of insight into real-world product experience. These publicly shared discussions offer an organic view of user expectations, frustrations, and success stories shaped by the specific contexts of use. Yet, harnessing this information for systematic analysis remains challenging due to the unstructured and domain-specific nature of the content. The lack of structure and specialized vocabulary makes it difficult for traditional data analysis techniques to accurately interpret, categorize, and quantify the feedback, thereby limiting its potential to inform product development and support strategies. To address these challenges, this paper presents the User eXperience Perception Insights Dataset (UXPID), a collection of 7130 artificially synthesized and anonymized user feedback branches extracted from a public industrial automation forum. Each JavaScript object notation (JSON) record contains multi-post comments related to specific hardware and software products, enriched with metadata and contextual conversation data. Leveraging a large language model (LLM), each branch is systematically analyzed and annotated for UX insights, user expectations, severity and sentiment ratings, and topic classifications. The UXPID dataset is designed to facilitate research in user requirements, user experience (UX) analysis, and AI-driven feedback processing, particularly where privacy and licensing restrictions limit access to real-world data. UXPID supports the training and evaluation of transformer-based models for tasks such as issue detection, sentiment analysis, and requirements extraction in the context of technical forums.

Choro Ulan Uulu, Mikhail Kulyabin, Layan Etaiwi et al.

Computer-Aided Engineering (CAE) enables simulation experts to optimize complex models, but faces challenges in user experience (UX) that limit efficiency and accessibility. While artificial intelligence (AI) has demonstrated potential to enhance CAE processes, research integrating these fields with a focus on UX remains fragmented. This paper presents a multivocal literature review (MLR) examining how AI enhances UX in CAE software across both academic research and industry implementations. Our analysis reveals significant gaps between academic explorations and industry applications, with companies actively implementing LLMs, adaptive UIs, and recommender systems while academic research focuses primarily on technical capabilities without UX validation. Key findings demonstrate opportunities in AI-powered guidance, adaptive interfaces, and workflow automation that remain underexplored in current research. By mapping the intersection of these domains, this study provides a foundation for future work to address the identified research gaps and advance the integration of AI to improve CAE user experience.

Hongyi Zhang, Zhiqiang Qi, Jingya Li et al.

Fast and reliable wireless communication has become a critical demand in human life. In the case of mission-critical (MC) scenarios, for instance, when natural disasters strike, providing ubiquitous connectivity becomes challenging by using traditional wireless networks. In this context, unmanned aerial vehicle (UAV) based aerial networks offer a promising alternative for fast, flexible, and reliable wireless communications. Due to unique characteristics such as mobility, flexible deployment, and rapid reconfiguration, drones can readily change location dynamically to provide on-demand communications to users on the ground in emergency scenarios. As a result, the usage of UAV base stations (UAV-BSs) has been considered an appropriate approach for providing rapid connection in MC scenarios. In this paper, we study how to control multiple UAV-BSs in both static and dynamic environments. We use a system-level simulator to model an MC scenario in which a macro BS of a cellular network is out of service and multiple UAV-BSs are deployed using integrated access and backhaul (IAB) technology to provide coverage for users in the disaster area. With the data collected from the system-level simulation, a deep reinforcement learning algorithm is developed to jointly optimize the three-dimensional placement of these multiple UAV-BSs, which adapt their 3-D locations to the on-ground user movement. The evaluation results show that the proposed algorithm can support the autonomous navigation of the UAV-BSs to meet the MC service requirements in terms of user throughput and drop rate.

Hongyi Zhang, Jingya Li, Zhiqiang Qi et al.

Fast and reliable connectivity is essential to enhancing situational awareness and operational efficiency for public safety mission-critical (MC) users. In emergency or disaster circumstances, where existing cellular network coverage and capacity may not be available to meet MC communication demands, deployable-network-based solutions such as cells-on-wheels/wings can be utilized swiftly to ensure reliable connection for MC users. In this paper, we consider a scenario where a macro base station (BS) is destroyed due to a natural disaster and an unmanned aerial vehicle carrying BS (UAV-BS) is set up to provide temporary coverage for users in the disaster area. The UAV-BS is integrated into the mobile network using the 5G integrated access and backhaul (IAB) technology. We propose a framework and signalling procedure for applying machine learning to this use case. A deep reinforcement learning algorithm is designed to jointly optimize the access and backhaul antenna tilt as well as the three-dimensional location of the UAV-BS in order to best serve the on-ground MC users while maintaining a good backhaul connection. Our result shows that the proposed algorithm can autonomously navigate and configure the UAV-BS to improve the throughput and reduce the drop rate of MC users.

Yuchu Liu, David Issa Mattos, Jan Bosch et al.

Randomised field experiments, such as A/B testing, have long been the gold standard for evaluating the value that new software brings to customers. However, running randomised field experiments is not always desired, possible or even ethical in the development of automotive embedded software. In the face of such restrictions, we propose the use of the Bayesian propensity score matching technique for causal inference of observational studies in the automotive domain. In this paper, we present a method based on the Bayesian propensity score matching framework, applied in the unique setting of automotive software engineering. This method is used to generate balanced control and treatment groups from an observational online evaluation and estimate causal treatment effects from the software changes, even with limited samples in the treatment group. We exemplify the method with a proof-of-concept in the automotive domain. In the example, we have a larger control ($N_c=1100$) fleet of cars using the current software and a small treatment fleet ($N_t=38$), in which we introduce a new software variant. We demonstrate a scenario that shipping of a new software to all users is restricted, as a result, a fully randomised experiment could not be conducted. Therefore, we utilised the Bayesian propensity score matching method with 14 observed covariates as inputs. The results show more balanced groups, suitable for estimating causal treatment effects from the collected observational data. We describe the method in detail and share our configuration. Furthermore, we discuss how can such a method be used for online evaluation of new software utilising small groups of samples.

Yuchu Liu, Jan Bosch, Helena Holmström Olsson et al.

A/B experimentation is a known technique for data-driven product development and has demonstrated its value in web-facing businesses. With the digitalisation of the automotive industry, the focus in the industry is shifting towards software. For automotive embedded software to continuously improve, A/B experimentation is considered an important technique. However, the adoption of such a technique is not without challenge. In this paper, we present an architecture to enable A/B testing in automotive embedded software. The design addresses challenges that are unique to the automotive industry in a systematic fashion. Going from hypothesis to practice, our architecture was also applied in practice for running online experiments on a considerable scale. Furthermore, a case study approach was used to compare our proposal with state-of-practice in the automotive industry. We found our architecture design to be relevant and applicable in the efforts of adopting continuous A/B experiments in automotive embedded software.

Yuchu Liu, David Issa Mattos, Jan Bosch et al.

A/B testing is gaining attention in the automotive sector as a promising tool to measure causal effects from software changes. Different from the web-facing businesses, where A/B testing has been well-established, the automotive domain often suffers from limited eligible users to participate in online experiments. To address this shortcoming, we present a method for designing balanced control and treatment groups so that sound conclusions can be drawn from experiments with considerably small sample sizes. While the Balance Match Weighted method has been used in other domains such as medicine, this is the first paper to apply and evaluate it in the context of software development. Furthermore, we describe the Balance Match Weighted method in detail and we conduct a case study together with an automotive manufacturer to apply the group design method in a fleet of vehicles. Finally, we present our case study in the automotive software engineering domain, as well as a discussion on the benefits and limitations of the A/B group design method.

David Issa Mattos, Lucas Ruud, Jan Bosch et al.

Benchmark suites, i.e. a collection of benchmark functions, are widely used in the comparison of black-box optimization algorithms. Over the years, research has identified many desired qualities for benchmark suites, such as diverse topology, different difficulties, scalability, representativeness of real-world problems among others. However, while the topology characteristics have been subjected to previous studies, there is no study that has statistically evaluated the difficulty level of benchmark functions, how well they discriminate optimization algorithms and how suitable is a benchmark suite for algorithm comparison. In this paper, we propose the use of an item response theory (IRT) model, the Bayesian two-parameter logistic model for multiple attempts, to statistically evaluate these aspects with respect to the empirical success rate of algorithms. With this model, we can assess the difficulty level of each benchmark, how well they discriminate different algorithms, the ability score of an algorithm, and how much information the benchmark suite adds in the estimation of the ability scores. We demonstrate the use of this model in two well-known benchmark suites, the Black-Box Optimization Benchmark (BBOB) for continuous optimization and the Pseudo Boolean Optimization (PBO) for discrete optimization. We found that most benchmark functions of BBOB suite have high difficulty levels (compared to the optimization algorithms) and low discrimination. For the PBO, most functions have good discrimination parameters but are often considered too easy. We discuss potential uses of IRT in benchmarking, including its use to improve the design of benchmark suites, to measure multiple aspects of the algorithms, and to design adaptive suites.

Hongyi Zhang, Jan Bosch, Helena Holmström Olsson

With the development and the increasing interests in ML/DL fields, companies are eager to apply Machine Learning/Deep Learning approaches to increase service quality and customer experience. Federated Learning was implemented as an effective model training method for distributing and accelerating time-consuming model training while protecting user data privacy. However, common Federated Learning approaches, on the other hand, use a synchronous protocol to conduct model aggregation, which is inflexible and unable to adapt to rapidly changing environments and heterogeneous hardware settings in real-world scenarios. In this paper, we present an approach to real-time end-to-end Federated Learning combined with a novel asynchronous model aggregation protocol. Our method is validated in an industrial use case in the automotive domain, focusing on steering wheel angle prediction for autonomous driving. Our findings show that asynchronous Federated Learning can significantly improve the prediction performance of local edge models while maintaining the same level of accuracy as centralized machine learning. Furthermore, by using a sliding training window, the approach can minimize communication overhead, accelerate model training speed and consume real-time streaming data, proving high efficiency when deploying ML/DL components to heterogeneous real-world embedded systems.

Lucy Ellen Lwakatare, Ellinor Rånge, Ivica Crnkovic et al.

Background: Data errors are a common challenge in machine learning (ML) projects and generally cause significant performance degradation in ML-enabled software systems. To ensure early detection of erroneous data and avoid training ML models using bad data, research and industrial practice suggest incorporating a data validation process and tool in ML system development process. Aim: The study investigates the adoption of a data validation process and tool in industrial ML projects. The data validation process demands significant engineering resources for tool development and maintenance. Thus, it is important to identify the best practices for their adoption especially by development teams that are in the early phases of deploying ML-enabled software systems. Method: Action research was conducted at a large-software intensive organization in telecommunications, specifically within the analytics R\&D organization for an ML use case of classifying faults from returned hardware telecommunication devices. Results: Based on the evaluation results and learning from our action research, we identified three best practices, three benefits, and two barriers to adopting the data validation process and tool in ML projects. We also propose a data validation framework (DVF) for systematizing the adoption of a data validation process. Conclusions: The results show that adopting a data validation process and tool in ML projects is an effective approach of testing ML-enabled software systems. It requires having an overview of the level of data (feature, dataset, cross-dataset, data stream) at which certain data quality tests can be applied.

Terese Besker, Antonio Martini, Jan Bosch

When developing software, it is vitally important to keep the level of technical debt down since it is well established from several studies that technical debt can, e.g., lower the development productivity, decrease the developers' morale, and compromise the overall quality of the software. However, even if researchers and practitioners working in today's software development industry are quite familiar with the concept of technical debt and its related negative consequences, there has been no empirical research focusing specifically on how software managers actively communicate and manage the need to keep the level of technical debt as low as possible.

Hugo Andrade, Ola Benderius, Christian Berger et al.

Heterogeneous computing is one of the most important computational solutions to meet rapidly increasing demands on system performance. It typically allows the main flow of applications to be executed on a CPU while the most computationally intensive tasks are assigned to one or more accelerators, such as GPUs and FPGAs. The refactoring of systems for execution on such platforms is highly desired but also difficult to perform, mainly due the inherent increase in software complexity. After exploration, we have identified a current need for a systematic approach that supports engineers in the refactoring process -- from CPU-centric applications to software that is executed on heterogeneous platforms. In this paper, we introduce a decision framework that assists engineers in the task of refactoring software to incorporate heterogeneous platforms. It covers the software engineering lifecycle through five steps, consisting of questions to be answered in order to successfully address aspects that are relevant for the refactoring procedure. We evaluate the feasibility of the framework in two ways. First, we capture the practitioner's impressions, concerns and suggestions through a questionnaire. Then, we conduct a case study showing the step-by-step application of the framework using a computer vision application in the automotive domain.

Hugo Andrade, Ivica Crnkovic, Jan Bosch

The most important way to achieve higher performance in computer systems is through heterogeneous computing, i.e., by adopting hardware platforms containing more than one type of processor, such as CPUs, GPUs, and FPGAs. Several types of algorithms can be executed significantly faster on a heterogeneous platform. However, migrating CPU-executable software to other types of execution platforms poses a number of challenges to software engineering. Significant efforts are required in such type of migration, particularly for re-architecting and re-implementing the software. Further, optimizing it in terms of performance and other runtime properties can be very challenging, making the process complex, expensive, and error-prone. Therefore, a systematic approach based on explicit and justified architectural decisions is needed for a successful refactoring process from a homogeneous to a heterogeneous platform. In this paper, we propose a decision framework that supports engineers when refactoring software systems to accommodate heterogeneous platforms. It includes the assessment of important factors in order to minimize the risk of recurrent problems in the process. Through a set of questions, practitioners are able to formulate answers that will help in making appropriate architectural decisions to accommodate heterogeneous platforms. The contents of the framework have been developed and evolved based on discussions with architects and developers in the automotive domain.

Nathan Bosch, Jan Bosch

System logs perform a critical function in software-intensive systems as logs record the state of the system and significant events in the system at important points in time. Unfortunately, log entries are typically created in an ad-hoc, unstructured and uncoordinated fashion, limiting their usefulness for analytics and machine learning. In this paper, we present the main challenges of contemporary approaches to generating and storing system logs data for large, complex, software-intensive systems based on an in-depth case study at a world-leading telecommunications company. Second, we present a systematic and structured approach for generating log data that does not suffer from the aforementioned challenges and is optimized for use in machine learning. Third, we provide validation of the approach based on expert interviews that confirms that the approach addresses the identified challenges and problems.

Jan Bosch, Ivica Crnkovic, Helena Holmström Olsson

Artificial intelligence (AI) and machine learning (ML) are increasingly broadly adopted in industry, However, based on well over a dozen case studies, we have learned that deploying industry-strength, production quality ML models in systems proves to be challenging. Companies experience challenges related to data quality, design methods and processes, performance of models as well as deployment and compliance. We learned that a new, structured engineering approach is required to construct and evolve systems that contain ML/DL components. In this paper, we provide a conceptualization of the typical evolution patterns that companies experience when employing ML as well as an overview of the key problems experienced by the companies that we have studied. The main contribution of the paper is a research agenda for AI engineering that provides an overview of the key engineering challenges surrounding ML solutions and an overview of open items that need to be addressed by the research community at large.

Anders Arpteg, Björn Brinne, Luka Crnkovic-Friis et al.

Surprisingly promising results have been achieved by deep learning (DL) systems in recent years. Many of these achievements have been reached in academic settings, or by large technology companies with highly skilled research groups and advanced supporting infrastructure. For companies without large research groups or advanced infrastructure, building high-quality production-ready systems with DL components has proven challenging. There is a clear lack of well-functioning tools and best practices for building DL systems. It is the goal of this research to identify what the main challenges are, by applying an interpretive research approach in close collaboration with companies of varying size and type. A set of seven projects have been selected to describe the potential with this new technology and to identify associated main challenges. A set of 12 main challenges has been identified and categorized into the three areas of development, production, and organizational challenges. Furthermore, a mapping between the challenges and the projects is defined, together with selected motivating descriptions of how and why the challenges apply to specific projects. Compared to other areas such as software engineering or database technologies, it is clear that DL is still rather immature and in need of further work to facilitate development of high-quality systems. The challenges identified in this paper can be used to guide future research by the software engineering and DL communities. Together, we could enable a large number of companies to start taking advantage of the high potential of the DL technology.