A knowledge-based approach to real-time systems

We select the OSAM* model for the knowledge base representation [2] and extend it with real time aspects.The reason for our selection of OSAM* is that it has versatile semantic associa­ tion types that increase the functionality of the system. OSAM* supports generalization, aggregation, interaction, cross product composition associations and follows the object oriented paradigm. It supports encapsulation and multiple inheritance mechanisms. The association types con­ trol the visibility among different entities of the application domain. OSAM* class definition integrates not only the structural and behavioral properties of similar objects, but also specifies the knowledge rule among those properties. The rule base is an integral part of the system and is homogeneously designed and is integrated with object classes. Real time aspects require the management of the critical situations that may arise in application domains within the specified time limit. Recently, real time systems deal with very large amount of data that require a more intelligent kernel to allow reasoning about application domain rules. Among other related works is the HiPAC project [1], that attempts to deal with timing constraints, and proposes contingency plans. POSTGRES provides alerters and triggers with which forward and backward reasoning mechanism can be realized. SYBASE supports situation action rules on database operations with restricted conditions and actions. Permission lo copy willioul fee all or pail of lliis material is granted provided that die copies are not made or distributed for direct com­ mercial advantage, die ACM copyright notice and the title of Hie publication and its dale appear, and notice Is given that copying Is by permission of the Association for Computing Machinery. To copy olltenvise, or to republish, requires a fee and/or specific per­ mission. Our work will be the embedding of real time aspects into the rules and operations of each class. The timing constraints to an operation are more specific than to a class. A crucial problem in real time systems is the execution of an urgent request while another less urgent request is running. The resolution of these conflicts is the basis for most scheduling algorithms that are designed to deal with a more intelligent real time system. This rule base design is built on some basic definitions, such as event, priority and hard-deadline. These properties of class rules as well as operation rules are inherited by subclassesOur algorithm computes a dynamic weighted priority for each re­ quest by taking into account the criticalness of the request, time needed for the completion of the request, and the time left for the re­ quest. Then the request with the highest dynamic priority takes precedence over others. The uniform representation of rules as classes in the design of the knowledge base as­ sists in removing redundancy and improving the efficiency of the KBMS design.