In modern systems engineering, domain experts increasingly utilize models to define domain-specific viewpoints in a highly interdisciplinary context. Despite considerable advances in developing model composition techniques, their integration in a largely heterogeneous language landscape still poses a challenge. Until now, composition in practice mainly focuses on developing foundational language components or applying language composition in smaller scenarios, while the application to extensive, heterogeneous languages is still missing. In this paper, we report on our experiences of composing sophisticated modeling languages using different techniques simultaneously in the context of heterogeneous application areas such as assistive systems and cyber-physical systems in the Internet of Things. We apply state-of-the-art practices, show their realization, and discuss which techniques are suitable for particular modeling scenarios. Pushing model composition to the next level by integrating complex, heterogeneous languages is essential for establishing modeling languages for highly interdisciplinary development teams.

Logging still is a core functionality used to understand the behavior of programs and executable models. Yet, modeling languages rarely consider logging as a first-level activity that is manifested in the language through modeling elements or their behavior. When logging is part of the code generated for the respective models or the corresponding runtime environment only, it must be generic, as the modeler cannot influence, through the models, what and when logging takes place. To enable modelers to log model behavior, we devised a method based on language extension and smart code generation that can integrate logging into arbitrary textual modeling languages. Based on this method, log entries can be produced, traced, and presented through a web application. This method and its infrastructure can facilitate lifting logging to the model level and, hence, improve the understanding of executable models.

Internet of Things (IoT) devices and the software they execute are often strongly coupled with vendors preinstalling their software at the factory. Future IoT applications are expected to be distributed via app stores. A strong coupling between hard- and software hinders the rise of such app stores. Existing model-driven approaches for developing IoT applications focus largely on the behavior specification and message exchange but generate code that targets a specific set of devices. By raising the level of abstraction, models can be utilized to decouple hard- and software and adapt to various infrastructures. We present a concept for a model-driven app store that decouples hardware and software development of product lines of IoT applications.

Agent-based simulation is an important testing tool for the development of autonomous vehicle software. Simulators enable engineers to test autonomous driving behavior in virtual environments, which is cheaper, faster, and safer than using a physical vehicle. An important aspect of autonomous driving software is its real-time capability, i.e. its ability to react to unforeseen events and new sensor inputs within a very short amount of time to prevent accidents. In this paper, we present a modular agent-based simulator architecture, which not only simulates the physical behavior of the vehicle, controlled by the software under test, but also its electrical/electronic (E/E) network. In particular, each ECU is simulated using a hardware emulator, which enables us to test the software as if it is run on the actual target hardware. Furthermore, the hardware emulator estimates the execution delays for the software under test, which enables more realistic approximations of the real behavior. In an evaluation example we analyze empirically how well the timing estimates reflect the reality. We show that modeling the memory hierarchy and instruction decoding has a crucial effect on the precision of this estimation.

Today’s Internet of Things (IoT) applications are mostly developed as a bundle of hardware and associated software. Future cross-manufacturer app stores for IoT applications will require that the strong coupling of hardware and software is loosened. In the resulting IoT applications, a quintessential challenge is the effective and efficient deployment of IoT software components across variable networks of heterogeneous devices. Current research focuses on computing whether deployment requirements fit the intended target devices instead of assisting users in successfully deploying IoT applications by suggesting deployment requirement relaxations or hardware alternatives. This can make successfully deploying large-scale IoT applications a costly trial-and-error endeavor. To mitigate this, we have devised a method for providing such deployment suggestions based on search and backtracking. This can make deploying IoT applications more effective and more efficient, which, ultimately, eases reducing the complexity of deploying the software surrounding us.

Machine learning is a discipline which has become ubiquitous in the last few years. While the research of machine learning algorithms is very active and continues to reveal astonishing possibilities on a regular basis, the wide usage of these algorithms is shifting the research focus to the integration, maintenance, and evolution of AI-driven systems. Although there is a variety of machine learning frameworks on the market, there is little support for process automation and DevOps in machine learning-driven projects. In this paper, we discuss how metamodels can support the development of deep learning frameworks and help deal with the steadily increasing variety of learning algorithms. In particular, we present a deep learning-oriented artifact model which serves as a foundation for build automation and data management in iterative, machine learning-driven development processes. Furthermore, we show how schema and reference models can be used to structure and maintain a versatile deep learning framework. Feasibility is demonstrated on several state-of-the-art examples from the domains of image and natural language processing as well as decision making and autonomous driving.

Developers questioning why their system behaves differently than expected often have to rely on time-consuming and error-prone manual analysis of log files. Understanding the behavior of Internet of Things (IoT) applications is a challenging task because they are not only inherently hard-to-trace distributed systems, but their integration with the environment via sensors adds another layer of complexity. Related work proposes to record data during the execution of the system, which can later be replayed to analyze the system. We apply the model-driven development approach to this idea and leverage digital twins to collect the required data. We enable developers to replay and analyze the system’s executions by applying model-to-model transformations. These transformations instrument component and connector (C&C) architecture models with components that reproduce the system’s environment based on the data recorded by the system’s digital twin. We validate and evaluate the feasibility of our approach using a heating, ventilation, and air conditioning (HVAC) case study. By facilitating the reproduction of the system’s behavior, our method lowers the barrier to understanding the behavior of model-driven IoT systems.

Digital twins emerge in many disciplines to support engineering, monitoring, controlling, and optimizing cyber-physical systems, such as airplanes, cars, factories, medical devices, or ships. There is an increasing demand to create digital twins as representation of cyber-physical systems and their related models, data traces, aggregated data, and services. Despite a plethora of digital twin applications, there are very few systematic methods to facilitate the modeling of digital twins for a given cyber-physical system. Existing methods focus only on the construction of specific digital twin models and do not consider the integration of these models with the observed cyber-physical system. To mitigate this, we present a fully model-driven method to describe the software of the cyber-physical system, its digital twin information system, and their integration. The integration method relies on MontiArc models of the cyber-physical system's architecture and on UML/P class diagrams from which the digital twin information system is generated. We show the practical application and feasibility of our method on an IoT case study. Explicitly modeling the integration of digital twins and cyber-physical systems eliminates repetitive programming activities and can foster the systematic engineering of digital twins.

Stations in Cyber-Physical Systems (CPSs) and especially Industrial Internet of Things applications often work towards a common goal, but not all their tasks may be equally important to reach this goal. Hence, stations need to prioritize traffic because network resources are limited. To increase the service quality by utilizing otherwise unused network resources, stations may also opt to use cooperation instead of contention. Many existing standards and academic approaches focus on implementing either cooperation or Quality of Service (QoS) mechanisms. In this paper, we evaluate how to leverage cooperation to improve QoS. Since stations in wireless industrial applications often communicate locally, we focus on the MAC layer. We identify a set of useful cooperation mechanisms that increase the packet delivery ratio, and then extend them by several packet prioritization strategies and evaluate in multiple simulated industrial scenarios. As a result, we provide a set of guidelines for protocol designers to combine different mechanisms depending on the requirements imposed by industrial applications. Moreover, we provide and evaluate an exemplary combination of mechanisms derived from our results aiming at high reliability and low latency.

Exhaustive testing of wireless communication protocols on prototypical hardware is costly and time-consuming. An alternative approach is network simulation, which, however, often strongly abstracts from the actual hardware. Especially in the wireless domain, such abstractions often lead to inaccurate simulation results. Therefore, we propose a code-transparent discrete event simulator that enables a direct simulation of existing code for wireless prototypes. With a focus on lower layers of the communication stack, we enable a parametrization of the simulation timings based on real-world measurements to increase the simulation accuracy. Our evaluation shows that we achieve close results for throughput (deviation below 3% for UDP and latency (corrected deviation about 13% compared to real-world setups, while providing the benefits of code-transparent simulation, i.e., to flexibly simulate large topologies with existing prototype code. Moreover, we demonstrate that our approach finds implementation defects in existing hardware prototype software, which are otherwise difficult to track down in real deployments.