Techniques and tools for model-based design and analysis of embedded real-time software.

As Embedded Real-Time (ERT) systems become more complex and safety-critical, there is a trend to raise software development level of abstraction from programming languages to models. We have developed a suite of techniques and tools to improve industry acceptance of model-driven development of ERT software. As a result of collaboration among multiple institutions, an end-to-end tool-chain has been developed for the design and analysis of ERT software, with Avionics Mission Computing (AMC) as the main target application. As part of the tool-chain, we have developed a tool called AIRES for model-level static analysis. Compared to traditional static analysis techniques that work at the level of programming languages, AIRES works at a higher level of abstraction, and provides valuable dependency and timing information to the engineer at an early stage of the design cycle. AIRES mainly focuses on the static structural aspects while largely ignoring the dynamic behavior of component interactions. We use model-checking to formalize the natural language description of the dynamic behavior of the AMC software, and verify safety and liveness properties. We also present several techniques to improve scalability of model-checking by exploiting application-level domain semantics. To bridge the gap between logical models and implementation on the physical execution platform, many UML tools come with automatic code generators that translate models into code in a programming language. However, current code generation technology generates functional code without considering non-functional and real-time issues. We have adapted the schedulability analysis algorithm by Harbour, Klein and Lehoczky to fit the native runtime model of UML-RT, a UML profile widely used in the telecom domain. This algorithm can be used during state-space exploration to synthesize an implementation architecture for a logical UML-RT model that satisfies timing constraints. In summary, the techniques and tools developed in this thesis address multiple aspects of model-driven development of ERT software, in order to shift the focus of the software development process from programming language-level to the model-level, and reduce the overall system development cost.Ph.D.Applied SciencesComputer scienceUniversity of Michigan, Horace H. Rackham School of Graduate Studieshttp://deepblue.lib.umich.edu/bitstream/2027.42/124418/2/3138164.pd