Parallel Discrete Event Simulation of Complex Propagation Phenomena : Parallel Interfaces and Cellular Environments

In this paper, we study principles and techniques of Parallel Discrete Event Simulation that are used for the simulation of environmental complex phenomena on parallel computers. The rise of new concepts in this domain allows to develop tools for the study of such systems. Among them, the paradigm of cellular automata facilitates the building of spatial models and the integration of parallel techniques. Some parallel programming interfaces are developed to accelerate the simulation treatments and exploit the computing ressources provided by the parallel computers. The Combination of the cellular automata and the parallel interfaces makes it possible to solve the modelling and simulation problems arisen by the complex systems considered. We present the modelling tools available for the expression of the space dynamics of the complex phenomenon considered and how to extend the possibilities of the traditional models for spatial modelling and simulation. Introduction Studying global ecosystem is nowadays essential to understand the human impact on such a fragile equilibrium. The modelling of environmental complex phenomena which occur in constitutes a necessary starting point. In fact, precise models only permit to design tools for a better understanding and to facilitate decision making. With the computing development, the simulation is becoming the best methodological choice possible to supply numerous solutions. However, the simulation of complex environmental phenomena often implies spatial effects which generate major problems in terms of memory and time computing. In addition the tools developed must predict the behavior of the phenomena in advance in order to operate the best way. Thus, as far as the simulation of forest fire propagation is concerned the development of a simulation environment requires the use of techniques able to apprehend the difficulties arisen by these issues. Parallel simulation has recently become an essential tool. Combined with the techniques of discrete events simulation and with the cellular automata paradigm, it makes it possible to solve the difficult problems related to the programming of such models. The parallelisation of these models depends on parallel programming interfaces. This document deals with the use of parallel interfaces and cellular automata in discrete event simulation. Thanks to this combination, it is possible to solve the problems aroused by the simulation of complex propagation phenomena such as vegetation fires. A first part presents the paradigms of the Parallel Discrete Event Simulation. A second part deals with the principal interfaces for parallel programming. A third part presents some environments which exploit parallelism based on cellular models. Eventually, a discussion starts the reflexion which lays the bases of the research tasks we plan to develop in order to model the propagation of vegetation fires. * Corresponding author