Design Space Exploration of incompletely specified Embedded Systems by Genetic Algorithms
GI/ITG/GMM Workshop "Methoden und Beschreibungssprachen zur Modellierung und Verifikation von Schaltungen und Systemen", 5.-7. April 2005, München
Abstract
Due to their growing complexity embedded systems are no longer designed from scratch. Therefore, a design methodology is necessary which allows to reuse existing submodules in combination with new modules, which feature a-priori unknown execution properties. This paper focuses on a genetic design space exploration algorithm, which jointly determines a complete set of Pareto optimal implementation alternatives (allocation, binding, and scheduling) including maximal allowed executions times for new modules in a single optimization run. This design space exploration method is based on a hierarchical specification model, which captures both data and control flow information. The figures of merit of the proposed approach are demonstrated by means of a real life example specification of a mobile robot.
Authors
Stephan Klaue, Ralf Laue, and Sorin A. HussDownload
Article (PDF)Slides (PPT)