PEABS: A Process for developing Efficient Agent-Based Simulators

Abstract

Multi-agent systems have been commonly used for implementing simulators in which several individuals or components perform actions. Although these systems usually manage large amounts of data, the process for constructing efficient multi-agent simulators is still an open issue. This paper presents a Process for developing Efficient Agent-Based Simulators, called PEABS, which combines the use of an agent-oriented modeling language with a framework adapted for obtaining efficient agent-based simulators. The proposed process covers all the development phases from the requirements elicitation until the deployment of a complete functional multi-agent system for simulation. As a case study, a simulator, which consists of an agent-based system that imitates tourist behaviors when selecting routes in the historic center of Madrid city, is constructed following the new process. The process is evaluated with this simulator and another one that simulates the evolution of student sociograms. These simulators with the same specifications were developed using an alternative process. The experimental comparison shows that the time devoted to development are at least 22% shorter with the current process. The systems constructed with this new process are more efficient in terms of execution time and memory usage than the systems developed with the other process. The final systems developed with PEABS are also more usable for the domain experts than the ones developed with the alternative process.