A Model-Driven Approach for Designing Industrial Embedded Systems

Abstract

Embedded systems are inherently complex being highly concurrent, reactive, and real-time. Consequently, embedded software development needs to consider both the functional and non-functional requirements of the system. Great emphasis is placed on the efficiency-oriented relationship between power consumption, available memory and execution speed. Many potential software development approaches exist, yet they often rely on using efficient programming languages like C in an ad hoc manner. As a result, the functional and non-functional requirements are met through the application of many low- level coding decisions which are tailored to a given hardware platform. However, such low-level approaches are time consuming, error prone, and difficult to maintain. They intertwine the specification with the underlying implementation, resulting in a problem that is further compounded by the fact that the requirements are never fully known at the outset. Model driven development (MDD) is being touted as a proposed methodology for alleviating these problems. Development starts with the specification of the system in a suitable semi-formal, and often visual, high-level language to guide the overall design. These high-level models are then subjected to rigorous analysis and testing before automated transformations enable code generation on a given target platform. Such platform-independent MDD has the potential to produce high-quality implementations, while significantly reducing the cost of maintenance. Tru-Test is a world leading embedded system developer in the Agri-Tech sector whose current design practice primarily relies on C. In this project we explored an alter- native approach, using a model driven standard for control systems called IEC 61499, to apply a more suitable computational model to a selected subsystem. The standard was also extended to include the benefits of the C and UML paradigms. This syntactic bridging between MDD and these popular paradigms provides a lower learning curve, reduced code duplication, improved modularity and more familiar options for designers.