The results described in this thesis form a part of the ongoing project: "A Simulation Environment for Laboratory Management by Robot Organizations" supported by NASA-Ames Cooperative Agreement No. NCC 2-525. Specific tasks dealt with in this thesis focus on the enhancement of DEVS-Scheme to the manipulation of continuous system models.
The long term goal is to create a tool for automatic transformation of continuous models into equivalent discrete event models. For this purpose, DEVS-Scheme must be able to execute continuous simulation runs, and generate time histories (trajectories) which can then be used by DEVS-Scheme to determine appropriate thresholds for switching from one state to another in an equivalent discrete event model.
DEVS-Scheme is a knowledge-based simulation environment for modeling and design. It facilitates the construction of families of models in a form that makes them easily reuseable by enabling the user to retrieve them from a model base. In DEVS-Scheme, the continuous system models have now become subclasses of discrete event models. These models provide the knowledge level descriptions of the continuous models. They also serve as pointers to corresponding continuous models that are stored in a continuous system model base called Dymola model base. The model structure and model dynamic descriptions are captured through Dymola models. Dymola is a hierarchically structured continuous system modelling language. By flattening the hierarchical structure of the model into a monolithic description, Dymola generates simulation programs. A compiler switch determines the target language. Previously, Dymola could either generate Simnon or Fortran models. As part of this MS Thesis, one additional target language, Desire, has been added. Desire is a direct-executing continuous system simulation language for efficient execution of continuous system simulation run. The simulated trajectories are then stored in a trajectory data base, which can later be referred to by DEVS-Scheme.
A System Entity Structure (SES) is a tree-structured knowledge representation scheme. It organizes all the knowledge about a system. The operation PRUNE generates a specific system among all the alternatives according to a design objective. If we TRANSFORM the pruned entity structure, the coupling continuous-system model will be generated in DEVS. At the same time, all the submodels needed for the system will be extracted from the Dymola model base, and the coupled Dymola system will be generated. Subsequently, TRANSFORM will call upon the Dymola preprocessor for generation of the executable Desire file. The RUN command finally will execute the Desire program which generates time histories to be stored in the trajectory data base.
As an example, the modelling and simulation of a fluid handling laboratory will illustrate how, in DEVS-Scheme, the SES is applied to manage the continuous models, and how simulation trajectaries are produced to form the time information needed by DEVS-Scheme for future reference.
Interested in reading the entire thesis? (160 pages, 19,099,648 bytes, pdf)