The developer community has made significant efforts to enable fast and resilient implementations of industrial applications centred around the Asset Administration Shell (AAS). Software Developer Kits (SDKs) help industry and academia to streamline the development process and reduce errors when working with the AAS metamodel or serving the AAS API. In this work, we analyze established SDKs and other supporting tools that facilitate the development of AAS SDKs and applications. We introduce four workflows with varying degrees of automation that leverage these tools, discussing both current and anticipated challenges related to code generation, along with effective mitigation strategies. These workflows are qualitatively compared regarding their implementation and management effort, required skills, correctness, as well as future-proof maintainability. Based on six study cases of real-world AAS implementations, we ascertain the extent to which these workflows are already being applied in practice.

This paper presents a novel approach for the semantic comparison of Asset Administration Shells (AAS), i.e., digital twins within the context of Industry 4.0. As digital twins gain importance in optimizing industrial processes, ensuring interoperability and accurate versioning of AAS becomes critical. We propose a method that focuses on object model-level comparisons rather than mere syntactic differences, enabling a more meaningful assessment of semantic equivalence. Our findings highlight the challenges associated with traditional comparison methods and suggest a comprehensive framework for enhancing AAS versioning and integrity verification.

Legacy assembly diagrams, often provided as machine-unreadable images, hinder integration into the Industry 4.0 (I4.0) ecosystem. We propose a multi-step pipeline leveraging Large Language Models as a single simple-to-operate tool to recognize parts, relationships, and global identifiers, segment parts in images, and structure the data into Asset Administration Shell as I4.0 digital twins. Targeted prompts are designed for each step and evaluated on 15 diverse real-world diagrams. Results show that while the LLM reliably recognizes parts and link identifiers, there are still some open challenges with relationship extraction and semantic segmentation. Despite these limitations, LLMs provide a viable tool for semi-automated digitalization. Our end-to-end pipeline thus enables seamless integration of legacy diagrams into I4.0 systems.

In the context of Industry 4.0, the automatic code and test generation from state diagrams embedded in specifications is a critical challenge for software correctness. In this paper we present an approach that leverages Large Language Models (LLMs) for the recognition of state diagrams to generate code and unit tests automatically. We compare the performance of LLMs with traditional computer vision models, highlighting the advantages of LLMs in terms of generalization and simplicity of setup. The results on two prominent industrial communication protocols, PROFINET and OPC UA, demonstrate the applicability of the approach, achieving significant reductions in manual effort and improving the accuracy of code and test generation.

As Industrie 4.0 (I4.0) technologies continue to advance, ensuring the reliability and robustness of emerging standards like Asset Administration Shells (AAS) becomes paramount. This paper introduces a comprehensive testing approach for reactive (a.k.a. Type 2) AAS implementations, addressing the challenges posed by their intricate and dynamic nature. Our methodology adopts a design-by-contract approach to formalize API behavior. We design the client as a test oracle, where code contracts extensively validate the AAS operations, their interactions and effects. The related test cases cover various operation chains, including superpaths, service specifications, and serialization modifiers. The objective of this approach is to enhance the reliability and effectiveness of testing in the evolving landscape of AAS Type 2 implementations.

In the dawn of Industry 4.0, a new era of unprecedented efficiency beckons, powered by the intricate web of interconnected machines, manufacturing sites, vendors, and consumers. The linchpin of this transformative journey is the Digital Twin, embodied in the Asset Administration Shell. By meticulously mirroring every facet of the real-world asset, the AAS offers a standardized portal for real-time data acquisition and proactive asset management. We present a Servicer Oriented Architecture for implementing Industry 4.0 Digital Twins using the Asset Administration Shell. We propose partitioning the AAS into a microservice network and formalise service definitions using gRPC and Protobuf. Our architecture allows for transparent service relocation near the asset, leveraging efficient gRPC communication. Additionally, We demonstrate that code generation tools streamline boilerplate generation and simplify the architecture update process.

The realization of Industrie 4.0 potential depends on the utilization of digital twins, commonly implemented as Asset Administration Shells. AAS provides a strong framework for describing assets and their functionalities, and is expected to become a key component of digital twin for embedded systems. In the fields of Industrial Cyber Physical Systems and Industrial Internet Of Things, AAS serves as a bridge between microcontroller and high-performance systems.Current Software Development Kits for AAS focus rather on cloud and PC applications. Building on top of transpilation-based approaches, we explore novel ways how to adapt the SDKs for AAS in order to bring them to industrial embedded systems. Our method is designed to streamline the development process in particular for micro-controller and Industrial Control Systems contexts.

Digital twins, the virtual representations of physical assets, processes or systems, are becoming increasingly important in cyber-physical systems. They are a foundational block of Industry 4.0, the movement to digitalize industrial processes.The Asset Administration Shell (AAS) has been established as the preferable model to capture interoperable digital twins in the context of Industry 4.0. As asset administration shells become more prevalent, so does the need for specialized software development kits (SDKs) to manage them.The intertwined Industry 4.0 value chains are highly dependent on interoperability. The SDKs for AAS are often the direct interface between the components, and the malfunctioning of the SDKs leads to the breakage of the value chain. It is therefore important that we have the tools to determine how individual SDKs behave, and detect when they malfunction. In this work, we implement an approach to assessing the maturity of the SDKs for managing asset administration shells, and perform a thorough survey and evaluation of the existing SDKs.

Industry 4.0 developers need high-quality libraries in their preferred programming languages to efficiently and ergonomically create, edit and exchange digital twins as Asset Administration Shell (AAS). Existing specifications of the AAS can not be directly translated into programming code. This leads to many, often manual, re-implementations of the AAS in different projects and programming languages. We present an approach for automatic generation of software libraries based on the formalized AAS meta-model. This shows that the development with our approach is 5× faster for a new language and 10× faster for a new version of the AAS meta-model, respectively.

In Industry 4.0, the digital twin is represented as an Asset Administration Shell. The Asset Administration Shell meta-model is used to develop this digital twin. A developer needs ready-made exchange formats and data schemas to be able to work efficiently with libraries in his programming languages. The existing representations of the metamodel in XML, JSON and RDF, in their current form, have been created manually by individual working group members. This process is error-prone and inconsistent in the enforcement of serialization rules, leading to many manual re-implementations of the Asset Administration Shell in each representation. We present an approach to automatically generate the different data schemas for information exchange between partners in the Industry 4.0 value chain, based on the intermediate representation of the formalized meta-model. The approach enables a practical and simple development process of and is scalable across different exchange formats and their future changes.

The digital twin, or more precisely the Asset Administration Shell, is the key element for interoperability in Industrie 4.0. Together with the asset, it forms the I4.0 component. For the development of I4.0 components, the availability of suitable software development tools is becoming increasingly relevant. However, it is not yet specified how these tools are to be systematically tested for correctness, compliance and conformity. These topics are the focus of this work. We show different established testing methods, evaluate their suitability and describe an approach for a capable test environment.

Design-by-contract as a programming technique is becoming popular in Python community as various tools have been developed for automatically testing the code based on the contracts. However, there is no sufficiently large and representative Python code base with contracts to evaluate these different testing tools. We present Python-by-contract dataset containing 514 Python functions annotated with contracts using icontract library. We show that our Python-by-contract dataset can be easily used by existing testing tools that take advantage of contracts. The demo video can be found at https://youtu.be/08wZN-xh6mY.

Industrie 4.0 maps the digital twin as the Asset Administration Shell. To develop this digital twin, the Asset Administration Shell meta-model is used. A developer needs high-quality libraries in their programming languages to build the Asset Administration Shell in efficient and ergonomic way. The existing representations of the meta-model are inconvenient for moving from the specification to realizations in software development. This leads to many, often manual, re-implementations of the Asset Administration Shell in each programming language. We present an approach to automatically generate libraries based on the intermediate representation of the meta-model. The approach allows for practical and simple development process of Industrie 4.0 components, and scales across programming environments and future changes in the meta-model.

Digitalization is forging its path in the architecture, engineering, construction, operation (AECO) industry. This trend demands not only solutions for data governance but also sophisticated cyber-physical systems with a high variety of stakeholder background and very complex requirements. Existing approaches to general requirements engineering ignore the context of the AECO industry. This makes it harder for the software engineers usually lacking the knowledge of the industry context to elicit, analyze and structure the requirements and to effectively communicate with AECO professionals. To live up to that task, we present an approach and a tool for collecting AECO-specific software requirements with the aim to foster reuse and leverage domain knowledge. We introduce a common scenario space, propose a novel choice of an ubiquitous language well-suited for this particular industry and develop a systematic way to refine the scenario ontologies based on the exploration of the scenario space. The viability of our approach is demonstrated on an ontology of 20 practical scenarios from a large project aiming to develop a digital twin of a construction site.

Large image datasets such as ImageNet or open-ended photo websites like Flickr are revealing new challenges to image classification that were not apparent in smaller, fixed sets. In particular, the efficient handling of dynamically growing datasets, where not only the amount of training data but also the number of classes increases over time, is a relatively unexplored problem. In this challenging setting, we study how two variants of Random Forests (RF) perform under four strategies to incorporate new classes while avoiding to retrain the RFs from scratch. The various strategies account for different trade-offs between classification accuracy and computational efficiency. In our extensive experiments, we show that both RF variants, one based on Nearest Class Mean classifiers and the other on SVMs, outperform conventional RFs and are well suited for incrementally learning new classes. In particular, we show that RFs initially trained with just 10 classes can be extended to 1,000 classes with an acceptable loss of accuracy compared to training from the full data and with great computational savings compared to retraining for each new batch of classes.

The number of digital images is growing extremely rapidly, and so is the need for their classification. But, as more images of pre-defined categories become available, they also become more diverse and cover finer semantic differences. Ultimately, the categories themselves need to be divided into subcategories to account for that semantic refinement. Image classification in general has improved significantly over the last few years, but it still requires a massive amount of manually annotated data. Subdividing categories into subcategories multiples the number of labels, aggravating the annotation problem. Hence, we can expect the annotations to be refined only for a subset of the already labeled data, and exploit coarser labeled data to improve classification. In this work, we investigate how coarse category labels can be used to improve the classification of subcategories. To this end, we adopt the framework of Random Forests and propose a regularized objective function that takes into account relations between categories and subcategories. Compared to approaches that disregard the extra coarse labeled data, we achieve a relative improvement in subcategory classification accuracy of up to 22% in our large-scale image classification experiments.

In recent years, large image data sets such as "ImageNet", "TinyImages" or ever-growing social networks like "Flickr" have emerged, posing new challenges to image classification that were not apparent in smaller image sets. In particular, the efficient handling of dynamically growing data sets, where not only the amount of training images, but also the number of classes increases over time, is a relatively unexplored problem. To remedy this, we introduce Nearest Class Mean Forests (NCMF), a variant of Random Forests where the decision nodes are based on nearest class mean (NCM) classification. NCMFs not only outperform conventional random forests, but are also well suited for integrating new classes. To this end, we propose and compare several approaches to incorporate data from new classes, so as to seamlessly extend the previously trained forest instead of re-training them from scratch. In our experiments, we show that NCMFs trained on small data sets with 10 classes can be extended to large data sets with 1000 classes without significant loss of accuracy compared to training from scratch on the full data.