Keywords: data model, domain models, requirement analysis, software engineering
Astract: The development of software tools is a collaborative process involving both the domain experts and the software engineers. This requires efficient communication considering different expertise and perspectives. Additionally, the two groups utilize language and communication tools in disparate ways. This, in turn, may lead to hidden misunderstandings in the requirement analysis phase and potentially result in implementation problems later on, that is difficult and costly to correct. In this paper, we demonstrate the above mentioned challenge via a use case from the tunneling domain. In particular, during the requirement analysis phase for a software capable of handling the data model of the subsoil. The domain experts in the field can best express the complexity of their domain by describing its artifacts, which in most cases are incomprehensible to the software engineers. We outline a method that interleaves requirement analysis and software modeling to enable an iterative increase of the accuracy and completeness of the information extracted from those artifacts and integrated into a flexible software model, which can produce testable software code automatically. Furthermore, we present a prototypical implementation of our method and a preliminary evaluation of the approach.

Keywords: tunneling, data capture, invoicing, digitalization, documentation

Keywords: Developing Economies, e-Government, e-Government Interoperability, Enterprise Architecture
Astract: The growth and uptake of e-government in developing economies is still affected by the interoperability challenge, which can be perceived as an orchestration of several issues that imply the existence of gaps in methods used for e-government planning and implementation. To a great extent, various counterparts in developed economies have succeeded in addressing the method-related gaps by developing e-government enterprise architectures, as blueprints for guiding e-government initiatives in a holistic and manageable way. However, existing e-government enterprise architectures are country-specific to appropriately serve their intended purpose, while enterprise architecture frameworks or methods are generic to accommodate several enterprise contexts. The latter do not directly accommodate the unique peculiarities of e-government efforts. Thus, a detailed method is lacking that can be adapted by developing economies to develop e-government enterprise architectures that fit their contexts. To address the gap, this article presents research that adopted a Design Science approach to develop an e-Government Enterprise Architecture Framework (EGEAF), as an explicit method for guiding the design of e-government enterprise architectures in a developing economy. EGEAF was designed by extending the Architecture Development Method of The Open Group Architecture Framework (TOGAF ADM) to address requirements for developing interoperable e-government solutions in a developing economy. EGEAF was evaluated using two scenarios in the Ugandan context, and findings indicate that it is feasible; its design is understandable to enable its adoption and extension to accommodate requirements for developing interoperable e-government solutions in other developing economies.

Keywords: Artificial intelligence, DMN, Enterprise modeling, Explainable AI, Machine learning, Model-driven engineering
Astract: The Decision Model and Notation (DMN) modeling language allows the precise specification of business decisions and business rules. DMN is readily understandable by business users involved in decision management. However, as the models get complex, the cognitive abilities of humans threaten manual maintainability and comprehensibility. Proper design of the decision logic thus requires comprehensive automated analysis of e.g., all possible cases the decision shall cover; correlations between inputs and outputs; and the importance of inputs for deriving the output. In the paper, the authors explore the mutual benefits of combining human-driven DMN decision modeling with the computational power of Artificial Intelligence for DMN model analysis and improved comprehension. The authors propose a model-driven approach that uses DMN models to generate Machine Learning (ML) training data and show, how the trained ML models can inform human decision modelers by means of superimposing the feature importance within the original DMN models. An evaluation with multiple real DMN models from an insurance company evaluates the feasibility and the utility of the approach.

Keywords: MSDKVS, EMF, metamodelling, modeltransformation, M3B, transformation bridge, MDSE, abstract syntax, Sirius, graphical concrete syntax
Astract: Powerful metamodeling frameworks for realizing Model-Driven Software Engineering (MDSE) exist, each having its own strengths, weaknesses, functionalities, programming languages and engineering communities. By combining their individual benefits, it would be preferable to establish interoperability between them, i.e. transforming metamodels and models created in one framework into equivalent (meta)models in other frameworks. Thus enabling the freedom of choosing the metamodeling platform without risking a lock-in effect.Two of the most well known metamodeling frameworks, which are well documented and available for free are the Eclipse Modeling Framework (EMF) and the Modeling SDK for Visual Studio (MSDKVS). This thesis gives detailed explanations on how to achieve interoperability between EMF and MSDKVS on two levels, namely the abstract and graphical concrete syntax levels. The platforms are compared, their features are mapped accordingly and a bidirectional transformation bridge on their metamodel and model layers is implemented. The approach will be extensively evaluated by transforming available metamodels and generated or manually created models thereof. The evaluation regarding the validity, executability and the expressiveness of the transformation bridge is done quantitatively and qualitatively.

Keywords: Building Information Modeling (BIM), Geometric Algorithms, 3D Modelling
Astract: This work introduces a method for integrating 2d construction documentation-level architectural details into a 3d conceptual model of a building to produce a detailed surface model. The goal is to generate geometry that can be built in the designated material using the appropriate standardized techniques. In the first step, we subject each 2d detail to manual 1d feature extraction to determine those shapes that have an influence on the 3d model. We also extract the sharp features of the 3d model to obtain the building’s skeleton, consisting of edge curves and corners. We perform an alignment of the feature collection obtained from each detail with the 3d skeleton, in accordance with the architectural design. Our goal is to build a 2-manifold with a boundary at each corner of the 3d skeleton. Spanning ruled surfaces between neighbouring corner manifolds completes the final surface model. In this work we focus on the algorithm for constructing the corner manifolds: After the alignment of the feature collections with the 3d skeleton is performed, we calculate a rich descriptor, based on geometric relationship functions, for each feature. In addition, we construct its adjacency graph, containing all other features whose descriptor will change in case this feature is discarded. We then apply simultaneous procedural contraction to all feature collections affecting the same corner of the 3d model. In each step a preservation score is calculated for all features, based on their descriptors. The feature with the lowest score is discarded and the descriptors and adjacency graphs for all others recalculated. This contraction produces ruled surface segments that are eventually stitched together into a 2-manifold with a boundary. We evaluated the algorithm by building a prototype in MatLab and testing it on 170 detail combinations.

Keywords: 5G, OpenAirInterface, URRLC, Edge Computing, MEC, Robotics, ROS
Astract: The novel mobile telecommunications standard 5G offers many improvements compared to its predecessor 4G. While from an end consumer perspective, the increase in bandwidth and download speed are the most crucial, the 5G use case Ultra-Reliable Low Latency Communications (URLLC) is the most important for the Internet of Things (IoT). The IoT consists of connected devices and sensors that exchange data with each other and central nodes. Often, the communication between the sensors and a central node must not exceed a low latency, e.g., 10ms. While URLLC reduces the latency in 5G, the link to a distant central node may introduce substantial latencies. A solution to solve this issue is to physically move the central node as close as possible to the connected 5G device. This is the main concept behind edge computing in 5G. As a mobile robot can be classified as one or more "things" in the IoT, it has similar requirements. For example, when a robot is localizing itself in a known or unknown environment, it relies on Light Detection and Ranging (LIDAR) scans to map the world. This procedure is called Simultaneous Localization and Mapping (SLAM). It is crucial that the measurements from the LIDAR sensor are processed with low latency, as delays can compromise the accuracy of the localization. However, Simultaneous Localization and Mapping (SLAM) is also computationally expensive. Therefore, it is highly desired and sometimes required to offload this computation to another computer, as robots are often hardware-constrained.The underlying hypothesis of this work is that it is possible to offload a SLAM algorithm to the edge of a 5G network without compromising functionality and quality of the localization. To achieve this, existing approaches to edge computing in 5G are analyzed and compared. Further, the open source 5G implementation OpenAirInterface is adapted to support the edge computing use case, which has not been considered in its initial design. Finally, an experiment is conducted which compares the accuracy and functionality of the SLAM algorithm in three different deployments: Local, WiFi and 5G edge. The results of this thesis show that offloading SLAM to the 5G edge is possible, whereas the accuracy and functionality of the localization and the produced map are slightly lower compared to the WiFi deployment. While much can still be optimized, this work lays the foundation for 5G-enabled mobile robotics, fully operated with open source software.

Keywords: conceptual modeling, accessibility, web modeling, model engineering, GLSP
Astract: According to the World Health Organization, the number of disabilities is increasing steadily. Additionally, as the internet has become a fundamental aspect of our daily lives, it is essential to ensure that all individuals, regardless of their abilities, have equal access to different information and services. Despite a growing effort to promote inclusion worldwide, there are still inaccessible tools and technologies in the field of computation. This results in unequal access to information systems and products for many people with disabilities. Consequently, accessibility should also be a key element, particularly in conceptual modeling, since web modeling tools have grown in popularity, as these tools have become essential in various fields, allowing professionals to work together, create, and build intricate web-based systems. In this regard, web accessibility is crucial for web modeling tools, as it ensures that users with disabilities can contribute fully to the modeling process and have equal chances to benefit.This thesis presents existing web accessibility issues in conceptual modeling research by identifying current research gaps and delineating a vision toward more inclusive, i.e., disability-aware conceptual modeling, based on a systematic review of the literature and current modeling tools. One key finding relates to a gap in research and tool support concerning physical disabilities. Based on these results, this thesis presents the first modeling prototype that can be used keyboard-only, thereby including users with physical disabilities to engage in conceptual modeling.

Keywords: testing, model engineering, web modeling, GLSP, modeling tools
Astract: Testing plays an important role in the quality assurance of software applications. It serves as a means to uncover potential issues introduced by code changes. As modern software systems become increasingly complex, manual testing practices often do not guarantee comprehensive coverage. Hence, automated testing has emerged as a solution to ensure thorough evaluation. This thesis delves into the domain of testing diagram editors, particularly focusing on the Graphical Language Server Platform (GLSP), which utilizes the Language Server Protocol, enabling developers to construct web modeling tools. In accordance, this thesis investigates the challenges of testing diagram editors. The investigation delves into the mechanics of browser automation and assesses the strengths and limitations of modern web testing frameworks such as Selenium, Cypress, and Playwright. This evaluation lays the groundwork for developing an extensible and maintainable testing framework tailored to GLSP. Within this context, the thesis addresses essential questions concerning how diagrams can be tested and interactions automated.For this reason, developing a solution that effectively addresses the unique demands of such tools is explored and encompasses various factors, including architectural and design principles, to form a foundation of an efficient and adaptable testing framework.To conclude, a comprehensive test suite is constructed. This suite spans a range of scenarios, each with different aspects of GLSP-based diagram editors to assess the testing framework’s capabilities in addressing those challenges.

Keywords: conceptual modeling, knowledge graph, modulaisation, genetic algorithm, model-driven engineering
Astract: Conceptual models are an integral part of information systems. They abstract concrete instances to display and emphasise specific aspects. These models grow and over time lead to monolithic artefacts and increased complexity. Modularisation is a process and technique to reduce complexity and make conceptual models more comprehensible by decomposing monolithic artefacts into smaller, understandable modular units. However, automatically finding a good modularisation solution, especially for any conceptual model, is difficult.Currently, there is a lack of a framework that can automatically modularise any conceptual model without any manual operation, where users can easily control configurations and extend the framework. Using the same framework or tool for different conceptual models, e.g. UML or BPMN models, is impossible. Most automatic modularisation approaches focus on specific conceptual models.This thesis aims to develop a complete modularisation framework, called GGMF, which stands for Generic Genetic Modularisation Framework, with a user interface. Users can customise parameters and specify objective criteria for modularisation solutions. It embeds the CM2KG framework for transforming conceptual models into knowledge graphs. A genetic algorithm receives the knowledge graph as an input and determines modularisation solutions. The framework generates the solutions as modularised knowledge graphs. The framework is designed to be extensible. Users can define custom objectives for the fitness function and replace the genetic encoding and operators with their own definitions.Finally, the work is evaluated by focusing on the configuration aspects, the impact of parameter changes, the comprehensibility of modularisation solutions and comparison with state-of-the-art modularisation-related techniques.