Development of a discrete event system simulator requires not only domain knowledge but also modeling technique. It is ideal that one person can expertise both on domain knowledge and on modeling technique. But in most case, a domain expert differs from a modeling expert. A domain expert provides functional requirements of the target system at an abstract level. Unlike the former case, a modeling expert needs much detailed descriptions of system behavior. Therefore, we need a means to lessen a gap between abstracted functional requirements and the detailed descriptions.
This thesis is aimed to propose a methodology for specification refinement at abstract levels, which can overcome the gap. To do so, this paper will first present the development of five levels of specification from the functional requirements to the description of system behavior. Then, it will explicate the description of each level while refining the information from the previous level.
The first level of the proposed methodology is functional requirements of the target system. The second level defines outer actors and usage scenarios. The third three defines system structure. The fourth level defines information exchange sequences among inner objects as a response to external inputs. UML (Unified Modeling Language) for development of a software system is used to specify the target system in steps the second level, the third level and the fourth level. Therefore UML is applied as a communication means between a domain expert and a modeling expert. Finally, the fifth level defines the detailed descriptions of system behavior at the discrete event system level. Therefore the final stage of the modeling will employ DEVS (Discrete Event System Specification) formalism, which specifies a discrete event system in a hierarchical, modular manner.
To demonstrate effectiveness of the proposed methodology, modeling and simulation of a concentration point/taxi simulator, a subsystem for a Call-Taxi system simulation, is exemplified.
