Rolls-Royce Control Systems supplies engine control and monitoring systems for aviation applications, and is required to design, certify, and deliver these to the highest level of safety assurance. To allow Rolls-Royce to develop safe and robust systems, which continue to increase in complexity, model-based techniques are now a critical part of the software development process. In this paper, we discuss the experiences, challenges and lessons learnt when developing a bespoke domain-specific modelling workbench based on open-source modelling technologies including the Eclipse Modelling Framework (EMF), Xtext, Sirius and Epsilon. This modelling workbench will be used to architect and integrate the software for all future Rolls-Royce engine control and monitoring systems.

Open-source model management frameworks such as OCL and ATL tend to focus on manipulating models built atop the Eclipse Modelling Framework (EMF), a de facto standard for domain specific modelling. MATLAB Simulink is a widely used proprietary modelling framework for dynamic systems that is built atop an entirely different technical stack to EMF. To leverage the facilities of open-source model management frameworks with Simulink models, these can be transformed into an EMF-compatible representation. Downsides of this approach include the synchronisation of the native Simulink model and its EMF representation as they evolve; the completeness of the EMF representation, and the transformation cost which can be crippling for large Simulink models. We propose an alternative approach to bridge Simulink models with open-source model management frameworks that uses an "on-the-fly" translation of model management constructs into MATLAB statements. Our approach does not require an EMF representation and can mitigate the cost of the upfront transformation on large models. To evaluate both approaches we measure the performance of a model validation process with Epsilon (a model management framework) on a sample of large Simulink models available on GitHub. Our previous results suggest that, with our approach, the total validation time can be reduced by up to 80%. In this paper, we expand our approach to support the management of Simulink requirements and dictionaries, and we improve the approach to perform queries on collections of model elements more efficiently. We demonstrate the use of the Simulink requirements and dictionaries with a case study and we evaluate the optimisations on collection queries with an experiment that compares the performance of a set of queries on models with different sizes. Our results suggest an improvement by up to 99% on some queries.

Conservative execution, end-to-end traceability, and context-aware resource handling are desirable features in model management build processes. Yet, none of the existing MDE-dedicated build tools (e.g. MTC-Flow, MWE2) support such features. An initial investigation of general-purpose build tools (e.g. ANT, Gradle) to assess whether we could build a workflow engine with support for these desirable features on top of it revealed limitations that could act as roadblocks for our work. As such, we decided to design and implement a new MDE-focused build tool (ModelFlow) from scratch to avoid being constrained by assumptions and technical constraints of these tools. We evaluated whether this decision was sensible by attempting to replicate its behaviour with Gradle in a typical model-driven engineering scenario. The evaluation highlighted scenarios where Gradle could not be extended to achieve the desirable behaviour which validates the decision to not base ModelFlow on top of it.

While the majority of research on Model-Based Software Engineering revolves around open-source modelling frameworks such as the Eclipse Modelling Framework, the use of commercial and closed-source modelling tools such as RSA, Rhapsody, MagicDraw and Enterprise Architect appears to be the norm in industry at present. This technical gap can prohibit industrial users from reaping the benefits of state-of-the-art research-based tools in their practice. In this paper, we discuss an attempt to bridge a proprietary UML modelling tool (PTC Integrity Modeller), which is used for model-based development of safety-critical systems at Rolls-Royce, with an open-source family of languages for automated model management (Epsilon). We present the architecture of our solution, the challenges we encountered in developing it, and a performance comparison against the tool’s built-in scripting interface. In addition, we use the bridge in a real-world industrial case study that involves the coordination with other bridges between proprietary tools and Epsilon.

MATLAB/Simulink is a tool for dynamic system modelling widely used across industries such as aerospace and automotive. Model management languages such as OCL, ATL and the languages of the Epsilon platform enable the validation, model-to-model, model-to-text transformation of models but tend to focus on the Eclipse Modelling Framework (EMF), a de facto standard for domain specific modelling. As Simulink models are built on an entirely different technical stack, the current solution to manipulate them using such languages requires their transformation into an EMF-compatible representation. This approach is expensive as (a) the cost of the transformation can be crippling for large models, (b) it requires the synchronisation of the native Simulink model and its EMF counterpart, and (c) the EMF-representation may be an incomplete copy of the model potentially hampering model management operations. In this paper we propose an alternative approach that uses the MATLAB API to bridge Simulink models with existing model management languages that relies on the "on-the-fly" translation of model management language constructs into MATLAB/Simulink commands. Our approach not only eliminates the cost of the transformation and of the co-evolution of the EMF-compatible representation but also enables full access to all the aspects of Simulink models. We evaluate the performance of both approaches using a set of model validation constraints executed on a sample of the largest Simulink models available on GitHub. Our evaluation suggests that the on-the-fly translation approach can reduce the model validation time by up to 80%.

In this paper we propose a domain specific language that enables the description and execution of model management workflows. Our language declares tasks and resources involved in a multi-step model management process and resolves the execution behaviour and order based on dependencies among these components. We describe the abstract and a concrete syntax of the language along with its execution semantics. Then, we demonstrate how the language interpreter can orchestrate and execute a selection model management tasks through a case study of a workflow that generates a graphical editor from a metamodel.

Mining data from remote repositories, such as GitHub and StackExchange, involves the execution of requests that can easily reach the limitations imposed by the respective APIs to shield their services from overload and abuse. Therefore, data mining clients are left alone to deal with such protective service policies which usually involves an extensive amount of manual implementation effort. In this work we present RestMule, a framework for handling various service policies, such as limited number of requests within a period of time and multi-page responses, by generating resilient clients that are able to handle request rate limits, network failures, response caching, and paging in a graceful and transparent manner. As a result, RestMule clients generated from OpenAPI specifications (i.e. standardized REST API descriptors), are suitable for intensive data-fetching scenarios. We evaluate our framework by reproducing an existing repository mining use case and comparing the results produced by employing a popular hand-written client and a RestMule client.