Portable and composable real-time software - a disruptive approach

ions in general-purpose programming languages Abstractions in embedded, hard real-time programming languages (5) Aspect-Oriented Programming (AOP) harnesses metaprogramming to improve the modularization of software (5) TDL-AOP for a better structuring of embedded software and avoidance of replicated code ) well-structured, maintainable code (4) domain-specific language extensions, for example, class=component libraries for Graphical User Interfaces (GUIs) (4) relevant aspects of control theory: 10=1 rule for actuator updates, advance calculation ) high-quality controllers with minimal hardware resources (3) object-orientation (abstract data types, inheritance, dynamic binding) (3) dynamic loading of TDL modules; instantiable TDL modules (e.g., for specifying redundancy in fault-tolerant systems); TDL mode extension through inheritance ) reusability of TDL components (2) modules (information hiding through abstract data structures) as means for component-based reuse and for structuring large software systems (2) TDL module ) transparent distribution (1) basic language abstractions and hierachical structuring constructs (functions and procedures) as foundation of high-level, platform-independent programming languages (1) Logical Execution Time (LET) ) determinism, portability originalarbeiten W. Pree Portable and composable real-time software – a disruptive approach heft 1–2.2007 | 11 Jänner/Februar 2007 | 124. Jahrgang tions and procedures set the stage of significanty more platformindependent programming in the late 1950s and 1960s. This ended the tedious machine level and assembly coding. (2) The module was invented as means to structure software in the large and to provide reusable software components. Modula and Ada as typical representatives of modular languages were defined in the 1970s and 1980s. (3) Object-oriented languages improved the reusability of modules. The class construct as instantiable module was originally provided by Simula in the 1960s. Smalltalk (1970s) and later on Cþþ (1980s), Java (1990s) and C# (2000) pushed object-orientation as common programming paradigm into the main stream. (4) Domain-specific extensions of languages have been delivered as class or component libraries since the 1980s. GUI (Graphical User Interface) libraries represent one important example of such a domain-specific extension. (5) So-called aspect-oriented programming (AOP), proposed in the 1990s, is heralded as post-object-oriented programming paradigm. It harnesses meta-programming to be able to modify the semantics of a programming language in order to improve the modularization of software. The items (1) and (2) in the right column of Table 1 summarize the achievements of the Giotto and MoDECS projects towards closing the gap between embedded and non-embedded software development: The pioneering LET abstraction delivers the important software property determinism and together with the TDL component model forms the foundation of transparent distribution (see section 4). TDLþ aims at delivering (3), (4) and (5) in the right column of Table 1. Radical innovation is required to invent and shape these missing abstractions: as (1) and (2) corroborate, the abstractions of high-level general purpose programming languages cannot be transferred directly to the real-time domain. For example, a module in general-purpose programming languages comprises other entities than a TDL module which has been defined according to the characteristics of hard real-time embedded systems. The same is true for the LET abstraction. The concepts and intentions of general-purpose programming language abstractions serve only as inspiration for the corresponding abstractions in the embedded hard real-time domain. 6.1 Expected results TDL has already been seamlessly integrated with Matlab=Simulink, which has been established as industry standard for modeling control applications. We accomplished the integration in the MoDECS project and called it the TDL:VisualCreator tool. This tool will be extended to comprise the TDLþ extensions. The fully automated code generation and the TDL run-time system guarantee that the behavior of a TDL module corresponds exactly to its specification. This is a characteristic and a principal advantage of model-based software development. The conventional way of showing the correctness of the code generators is by testing various sample applications. Instead of just testing the code generators we aim at verifying the most critical generator tool in the TDLþ project, the bus schedule generator. This represents another significant contribution to full-fledged model-based software development. To sum up, the principal goals of the planned future research activities are the following: " full-fledged model-based development by inventing and implementing abstractions (3), (4), and (5), resulting in the modeling language TDLþ; some of the control engineering enhancements (4) have already been accomplished " corresponding extensions of the MatLab=Simulink integration, that is, the TDLþ:VisualCreator tool " verification of the bus schedule generator 6.2 Practical relevance The TDL tools are currently used and evaluated by automotive suppliers and OEMs, in particular in the realm of time-triggered distributed systems development with FlexRay and with EtherCAT. TDL will also be used in the realm of the 2007 DARPA Urban Grand Challenge in the vehicle developed by the University of California, Berkeley. 7. Conclusion The LET abstraction invented in the realm of the Giotto project paved theway for transparent distribution in real-time systems. From our own experience and the feedback from various industry partners we are convinced that this novel approach is a breakthrough that will lead to significantly more robust embedded software and will at the same time reduce the costs of development and integration testing. Acknowledgements We thank the MoDECS project team at the University of Salzburg for providing valuable input during informal discussions and group meetings. This research was supported in part by the FIT-IT Embedded Systems grant 807144 provided by the Austrian government through its Bundesminsterium für Verkehr, Innovation und Technologie.