Model Driven Development for Embedded Systems

It is important for developing enterprise systems to fully analyze at an early stage the business workflows that describe interactions involving systems and their users. This is also important when developing embedded systems, which consist of various hardware components, their environment, and control software. Many combinations of sensors and actuators can be used to implement the requirements, and the control software will be different for every combination. At the requirements analysis phase, it is difficult to adequately test all possible combinations of hardware components. Recently, MDD (Model Driven Development) has become a promising approach for system development. Many researchers actively apply UML (Unified Modeling Language) to embedded systems, and model transformation is expected to determine the best combinations of modeling elements that depend on both the hardware architecture and the system environment. Executable UML is a key technology for expressing application domains in a platform-independent manner with formal action semantics using techniques such as class diagrams and state machine diagrams. This paper proposes a development method for embedded systems based on MDD in which the models are executable and testable. The effectiveness of our method is demonstrated through the development of a maze robot.