Open modeling and simulation framework for evolutive analysis

• Autores: De Giusti, Marisa Raquel; Lira, Ariel Jorge; Villarreal, Gonzalo Luján
• Año de publicación: 2008
• Idioma: inglés
• Tipo de recurso: documento de conferencia
• Materia: Ciencias de la Computación e Información; discrete event simulation; storage of runs; statistical and evolutive analysis; simulation teaching
• Institución: Comisión de Investigaciones Científicas de la Provincia de Buenos Aires

Simulation is the process of executing a model, that is a representation of a system with enough detail to describe it but not too excessive. This model has a set of entities, an internal state, a set of input variables that can be controlled and others that cannot, a list of processes that bind these input variables with the entities and one or more output values, which result from the execution of events.\nRunning a model is totally useless if it can not be analyzed, which means to study all interactions among input variables, model entities and their weight in the values of the output variables. In this work we consider Discrete Event Simulation, which means that the status of the system variables being simulated change in a countable set of instants, finite or countable infinite.\nExisting GPSS implementations and IDE's provide a wide range of tools for analysis of the results, for generation and execution of experiments and to perform complex analysis (such as Analysis of Variance, screening and so on). This is usually enough for most common analysis, but more detailed information and much more specific analysis are often required.\nIn addition, teaching this kind of simulation languages is always a challenge, since the way it executes the models and the abstraction level that their entities achieve is totally different compared to general purpose programming languages, well known by most students of the area. And this is usually hard for students to understand how everything works underground.\nWe have developed an open source simulation framework that implements a subset of entities of GPSS language, which can be used for students to improve the understanding of these entities. This tool has also the ability to store all entities of simulations in every single simulation time, which is very useful for debugging simulations, but also for having a detailed history of all entities (permanents and temporary) in the simulations, knowing exactly how they have behaved in every simulation time.\nIn this paper we provide an overview of this development, making special stress on the simulation model design and on the persistence of simulation entities, which are the basis that students and researchers of the area need in order to extend the model, adapt it to their needs or add all analysis tools to the framework.