| Research into real-time systems is for the most part dedicated to improving scheduling methods. As real-time systems grow in complexity, so also do the demands placed on their schedulers. It is expected that conventional scheduling algorithms will be inadequate to meet the needs of the next generation of real-time schedulers, requiring a fresh approach to the problem. Fuzzy-logic-based schedulers show promise as an alternative because of their flexibility, graceful performance degradation on transient overloads, and fault tolerance. However, actual implementations of fuzzy schedulers have hitherto suffered from an unacceptably high scheduling overhead. This thesis presents a method to implement a fuzzy priority scheduler with a low overhead using an efficient knowledge compilation scheme. A new modeling paradigm suitable for real-time systems, and an automatic program generator employing this modeling paradigm to generate real-time programs from a dataflow model of the system are also presented. |
