Enhancing a multi-agent system's performance: from implementation to simulation analysis

In a multi-agent system (MAS), agents preserve their proactivity and autonomy. Consequently, one associates an independent process with each of them, but when the simulation is only on a single mono-processor computer, the performance tends to be affected because computer resources are more limited. Unlike a system of multicomputers (where generally the real parallelism may be found), the management of these processes is more difficult in a mono-processor environment. We show that it is nonetheless possible to enhance the MAS performance in such an environment and discuss the way we do it. We determine the factors to be taken into account and refer to our GEAMAS platform in carrying out our investigation.