No abstract

No abstract

Discrete Event System Specification (DEVS) is a modular and hierarchical formalism for system modeling and simulation. DEVS models can be mathematically described; simulation is performed by tools called concrete simulators. Concerning atomic DEVS models, each concrete simulator has its own input language which is, essentially, a general-purpose programming language (such as Java or C+ +). Hence, once engineers have written the mathematical model, they need to manually translate it into the input language of the concrete simulator of their choice. In this paper we present a multitarget compiler for atomic DEVS models written in CML-DEVS, a mathematics-based DEVS modeling language. This multi-target compiler is able to compile a CML-DEVS model to the input languages of the PowerDEVS and DEVS-Suite concrete simulators. In this way, the CML-DEVS compiler frees engineers from the manual translation of their mathematical models. In fact, the same mathematical model can be simulated on both simulators by simply recompiling the model. The CML-DEVS multi-target compiler can be easily extended to produce code for other concrete simulators.

The most common method to validate a DEVS model against the requirements is to simulate it several times under different conditions, with some simulation tool. The behavior of the model is compared with what the system is supposed to do. The number of different scenarios to simulate is usually infinite, therefore, selecting them becomes a crucial task. This selection, actually, is made following the experience or intuition of an engineer. Here we present a family of criteria to conduct DEVS model simulations in a disciplined way and covering the most significant simulations to increase the confidence on the model. This is achieved by analyzing the mathematical representation of the DEVS model and, thus, part of the validation process can be automatized.