https://www.mdu.se/

The collaborative nature of model-driven software engineering introduces significant challenges in safeguarding the confidentiality and integrity of the collaborative model. Existing access control mechanisms often rely on transient, virtual views lacking persistence and fine-grained permissions, making them unsuitable for scenarios requiring offline collaboration and leading to potential security breaches and user frustration. This work describes a dual-layered approach leveraging role-based access control policies to enhance security in collaborative modeling environments. The first layer utilizes multi-view modeling techniques to create materialized view models tailored to specific user roles, thereby restricting unnecessary access to the entire model. The second layer refines access at the individual element level within these view models, establishing fine-grained permissions enforced by model editors. This proactive enforcement prevents unauthorized actions before they occur, improving user experience and efficiency. The proposed approach, implemented as an Eclipse plugin and demonstrated through an illustrative example, ensures the confidentiality and integrity of shared model data by granting stakeholders access only to information relevant to their specific responsibilities and expertise. By filtering out irrelevant data, the approach also mitigates information overload, enabling stakeholders to concentrate on task-relevant aspects of the model, thereby potentially improving collaborative efficiency and effectiveness. 

Software-intensive systems are characterized by intricate interactions, distributed structures, and ongoing evolution posing notable challenges for traceability, change analysis, and decision support. Modeling and versioning practices are widely used to manage this complexity; however, in practice, they are often fragmented, limiting the value that organizations can derive from them. To address this challenge, we map modeling and versioning practices to the benefits they enable, providing a structured framework for practitioners to assess their current practices and identify opportunities for incremental improvement. Building on findings from our previous interview study with multiple companies, together with insights from the literature, we identify four levels of modeling formality, five levels of versioning sophistication, and four categories of benefits, summarized in two mapping tables. Our contributions include a practitioner-oriented roadmap that helps teams reflect on current modeling and versioning practices, anticipate achievable benefits, and identify incremental adjustments that can add value with minimal overhead. An initial validation with industry representatives confirms the practical relevance of the roadmap and highlights its potential to guide structured reflection and improvement of modeling and versioning practices. 

Informal diagrams created in general-purpose diagramming tools are widely used in software architecture because they are quick to produce and easy to share. However, the lack of constraints in such tools often yields inconsistent notations and ad-hoc conventions, which in turn invite misinterpretation when diagrams are read outside their original context. Dedicated modeling languages and environments can mitigate these issues but are frequently resisted due to steep learning curves and disruptive adoption costs. Building on the flexible modeling paradigm, this paper proposes an approach to the lightweight formalization of informal diagrams. Grounded in observed industrial challenges and prior work on flexible modeling, we derive a set of design principles and instantiate them in an approach realized as a Draw.io plugin. We propose an approach that addresses challenges from industrial settings by enabling practitioners to introduce structure incrementally into the informal diagrams they already create, thereby helping resolve notational inconsistency and clarify meaning. Moreover, we show how our approach satisfies the guiding flexible modeling principles by introducing these model-like benefits without compromising the accessibility and speed of informal diagramming as valued in practice. The contribution enhances clarity and consistency within familiar workflows and lays the groundwork for subsequent capabilities such as rapid dissemination of conventions, enterprise-level aggregation, and additional quality checks should organizations choose to adopt them. © 2026, Science and Technology Publications, Lda. All rights reserved.

In this article, we discuss how we can facilitate the growing need for early validation and verification (V&V) of system behaviour in Model-Based Systems Engineering (MBSyE). Several aspects, such as reducing cost and time to market, push companies towards integration of V&V methods earlier in development to support effective decision-making. One foundational methodology seeing increased attention in industry is the use of MBSyE, which brings benefits of models with well-defined syntax and semantics to support V&V activities, rather than relying on natural language text documentation. Despite their promise, industrial adoption of these practices is still challenging. This article presents a vision for readily available early V&V. We present a summary of the literature on early V&V in MBSyE and position existing challenges regarding potential solutions and future investigations towards this vision. We elaborate our vision by means of challenges with a specific emphasis on early V&V of system behaviour. We identify three specific challenge areas: Creating and managing Models, Organisational systems engineering aspects, and early V&V Methods. Finally, we outline a road-map to address these categories of challenges, in which we propose the transfer of established best practices from the software engineering domain to support emerging technologies in the systems engineering domain. 

The increasing demand for autonomous machines in construction environments necessitates the development of robust object detection algorithms that can perform effectively across various weather and environmental conditions. This paper introduces a new semantic segmentation dataset specifically tailored for construction sites, taking into account the diverse challenges posed by adverse weather and environmental conditions. The dataset is designed to enhance the training and evaluation of object detection models, fostering their adaptability and reliability in real-world construction applications. Our dataset comprises annotated images captured under a wide range of different weather conditions, including but not limited to sunny days, rainy periods, foggy atmospheres, and low-light situations. Additionally, environmental factors such as the existence of dirt/mud on the camera lens are integrated into the dataset through actual captures and synthetic generation to simulate the complex conditions prevalent in construction sites. We also generate synthetic images of the annotations including precise semantic segmentation masks for various objects commonly found in construction environments, such as wheel loader machines, personnel, cars, and structural elements. To demonstrate the dataset’s utility, we evaluate state-of-the-art object detection algorithms on our proposed benchmark. The results highlight the dataset’s success in adversarial training models across diverse conditions, showcasing its efficacy compared to existing datasets that lack such environmental variability.

Models and metamodels change, entailing efforts to keep related artifacts consistent, i.e., to reflect the implications of the changes on them. In order to assess these implications, the changes or evolution steps themselves are, in most cases, of highest interest, compared to the states of the models or metamodels. While the states can be used to derive the changes, some information on the actual changes might get lost, e.g., whether an empty class has been renamed or deleted and re-added. The use of deltas to describe changes is not limited to models and metamodels, but is also employed in other research areas. To get an overview of the used concepts and how they compare, we did a scoping literature review in the field of computer science, focused on modeling and related fields. We compared the different approaches in regard to how they model the change, their ability to model atomic or composite changes, their completeness in modeling all possible changes, as well as their purpose. This overview allows for more efficient concept re-use across domains in regard to the modeling of changes and the different use cases realized with them.url: https://doi.ieeecomputersociety.org/10.1109/MODELS-C68889.2025.00065

Large language models have shown promising potential in reducing the accidental complexity of model-driven engineering, particularly in automating model transformation tasks. However, existing research has shown that large language models often struggle with correctness, generalizability, and performance in complex transformation scenarios. This paper investigates the feasibility of applying large language models within the paradigm of model transformation by example, an intuitive technique that derives transformation logic from prototypical source-target model pairs, thereby reducing cognitive load and bypassing the need for formal transformation specifications. We empirically benchmark five LLMs, ChatGPT-4.5, DeepSeek V3, DeepHermes 3 LLaMA 3 8B, QwQ 32B, and OlympicCoder 32B, across three transformation scenarios of increasing complexity: RDBMS-to-UML, UML-to-Java, and SysML-to-AAS. Each scenario is evaluated under three configurations, varying the number of example pairs provided. In each configuration, the LLMs are tasked with directly generating target models given source models, example pairs, and initial mapping rules. Model outputs are assessed using correctness and weighted success metrics that account for structural and semantic fidelity. Our findings reveal that LLMs perform well in syntactically regular transformations but struggle in semantically rich or complex scenarios. Additional examples do not consistently enhance performance and may even degrade it, highlighting the models’ limitations in abstraction and semantic grounding. GPT-4.5 delivers the highest peak performance but suffers from instability, while smaller models like DeepSeek V3 and OlympicCoder 32B offer more stable, scalable results.

This paper presents a systematic mapping study classifying existing scientific contributions on synergies of Model Driven Engineering (MDE), Artificial Intelligence/Machine Learning (AI/ML), and DevOps, with the overall objective of supporting the continuous development of Cyber-Physical Systems (CPSs). We collected papers from bibliographic sources and selected primary studies to analyse. Then, we characterised and classified the current state of the art, focusing on 1) main aspects already tackled at the intersection of at least two of the three studied areas, and 2) findings emerging from the analysis as a framework for potential future research, notably regarding the integration of the three studied areas. The results reveal that few approaches combine MDE, AI/ML, and DevOps for software and systems engineering. In contrast, several approaches have combined two of them, specifically MDE and DevOps. Approaches combining AI/ML with MDE or DevOps are also becoming more frequent and will most likely continue to progress in the future. These synergies cover a range of engineering activities, from requirements and design to monitoring, maintenance, and evolution. Open research challenges include advancing AI/ML, MDE, and DevOps integration, supporting scalable, data-oriented solutions, proposing new continuous engineering methods, and adapting DevOps practices to diverse systems.

Model-driven engineering envisions an enhancement of software engineering by promoting automation through model transformations. However, the effective use of model-driven tools often requires significant expertise due to their reliance on custom domain-specific languages for transformations. This expertise gap, combined with challenges like inadequate tool support and the need for additional training, has meant that model-driven engineering sometimes struggled to reduce, and might have even increased, accidental complexity. Addressing this problem, our work investigates the use of large language models, specifically ChatGPT-4, to reduce accidental complexity in model transformation processes within model-driven engineering. We conducted a systematic literature review and designed an experiment to explore ChatGPT-4's efficacy in performing model transformations out-of-the-box. Using a semi-automated pipeline, we applied ChatGPT-4 to 99 UML class diagram models, generating Java programs and comparing them with ground truth programs created by a state-of-the-art modelling tool. Our findings indicate a cumulative success rate of 94% after three iterations, with most generation errors being resolved during the process. However, complex models presented a significant challenge, with a cumulative success rate of only 17%.