A knowledge-based environment for hierarchical modelling and simulation.

Abstract

Hierarchical, modular specification of discrete-event models offers a basis for reusable model bases and hence for enhanced simulation of truly varied design alternatives. This dissertation develops a knowledge-based environment for hierarchical modelling and simulation of discrete-event systems as the major part of a longer, ongoing research project in artificial intelligence and distributed simulation. In developing the environment, a knowledge representation framework for modelling and simulation, which unifies structural and behavioral knowledge of simulation models, is proposed by incorporating knowledge representation schemes in artificial intelligence within simulation models. The knowledge base created using the framework is composed of a structural knowledge base called entity structure base and a behavioral knowledge base called model base. The DEVS-Scheme, a realization of DEVS (Discrete Event System Specification) formalism in a LISP-based, object-oriented environment, is extended to facilitate the specification of behavioral knowledge of models, especially for kernel models that are suited to model massively parallel computer architectures. The ESP-Scheme, a realization of entity structure formalism in a frame-theoretic representation, is extended to represent structural knowledge of models and to manage it in the structural knowledge base. An advantage of the knowledge-based environment is that it is capable of automatically synthesizing hierarchical, modular models from model base resident components defined by the extended DEVS-Scheme under the direction of structural knowledge using the extended ESP-Scheme. Since both implementation and the underlying LISP language are accessible to the user, the result is a medium capable of combining simulation modelling and artificial intelligence techniques. To show the power of the environment, modelling and simulation methodology in the environment are presented using an example of modelling a hypercube computer architecture. Applications of the environment to knowledge-based computer systems design, communications network design, and diagnostic expert systems design are discussed. Since structure descriptions in the environment are susceptible to run-time modification, the environment provides a convenient basis for developing variable family and variable structure simulation models such as adaptive computer architectures. Thus, the environment represents a significant step toward realizing powerful concepts of system-theoretic based formalisms. The environment also serves as a medium for developing distributed simulation architectures for hierarchical, modular discrete-event models.