No abstract

Model‐based design is a promising technique to improve the quality of software and the efficiency of the software development process. We are investigating how to efficiently model embedded software and its environment to verify the requirements for the system controlled by the software. The software environment consists of mechanical, electrical and other parts; modelling it involves learning how these parts work, deciding what is relevant to model and how to model it. It is not possible to fully automate these steps. There are general guidelines, but given that every modelling problem differs, much is left to the modeller's own preference, background and experience. Still, when the next generation of a system is designed, the new system will have common elements with its previous version. Therefore, lessons learned from the current model could inform future models. We propose a framework for identifying the non‐formal elements of knowledge, insights and a model itself, which can support modelling of the next system generation. We will present the application of our framework on an action research case – modelling mechanical parts of a paper‐inserting machine.

No abstract

As today's devices, gadgets and machines become more intelligent, the complexity of embedded software controlling them grows enormously. To deal with this complexity, embedded software is designed using model-based paradigms. The process of modelling is a combination of formal and creative, design steps. Because of the partially non-formal character of modelling, the relation between a model and the system cannot be expressed mathematically. Therefore, the modeller's justification that the model represents the system adequately can only be non-formal. In this paper we discuss the nature of non-formal modelling steps and pinpoint those that create a 'link' between the model and the system. We propose steps to structure the explanation and justification of non-fomal modelling decisions. This in turn should enhance confidence that the non-formal, physical world surrounding the embedded system is adequately represented in the model.

Researchers make a significant effort to develop new modelling languages and tools. However, they spend less effort developing methods for constructing models using these languages and tools. We are developing a method for building an embedded system model for formal verification. Our method provides guidelines to build a model and to construct a correctness argument. We start from a high-level formula stating that a plant (a device that performs a task) and its control should satisfy requirements. As our knowledge about the system grows, we refine this formula and the model gradually, in a stepwise non-monotonic process, until we have a description that can be formally verified. In this paper we explain our method on a simple example and compare it briefly with two other methods: requirements progression and the goal-oriented KAOS approach. The requirements progression is an extension of a problem frames approach. The KAOS method is also based on problem frames, but introduces new concepts for describing a system.