|
Authors: | Hans Otto Trutmann |
Group: | System Engineering |
Type: | PhD Thesis |
Title: | Separate Connection and Functionality is the Pivot in Embedded System Design |
Year: | 2000 |
Month: | October |
Pub-Key: | Tru00b |
Keywords: | Code Generation, Embedded Systems, Real-Time Systems, Software Specification |
ISBN: | 3-906469-09-3 |
ETH Nbr: | 13891 |
Pub Nbr: | 39 |
School: | ETH Zurich |
Abstract: | Aside from a solution that satisfies the initial requirements, development of embedded real-time systems is also concerned with appropriate descriptions of a system?s different aspects, as well as with connecting these representations among each other and with the implementation. Thus, the solution appears as a by-product of a well-organized development process. This process profits from a clear separation of concerns, which can be achieved with an explicit interconnection model, the architectural pivot specification. Its relative stability allows development to progress independently behind stable syntactic and semantic interfaces. The three basic activities in control applications, input, processing and output, are dealt with individually within separate problem areas. A problem-oriented approach postulates the use of suitable methods and tools for solving each of these issues, thus embodying different formal representations and explicitly written code on different levels of abstraction. Unison is achieved through generators yielding code with defined generic properties that favor smooth and efficient implementation. The functional problem comprises all the behavioral aspects of a control system, using shared phenomena to interact with the controlled processes in the environment. Solutions to this problem omit any details not pertaining to the relation between the course of events in the environment and the inner states of a model. The connection problem is concerned with information transport from and to the environment. Its solution is based on regular sequential structures. During development, the common high-level source baseline permits generating implementations for various modes of operation without modifying the abstract descriptions. The approach eliminates manual programming when code alterations are needed to adapt to changes in the hardware environment. It thus allows testing of proposed solutions against numerical models, as well as adapting resources according to emerging requirements any time during the development process. It supports decentralized development work and simplifies target implementations without complicated scheduling, without interrupts, and without the overhead of elaborate operating systems. The code is statically deterministic, and the performance of entire implementations is analyzed exhaustively to assure the desired real-time properties. |
Remarks: | Prüfung am 23.10.00 |
Resources: | [BibTeX] [Paper as PDF] |