High Level Architecture For Distributed Agent Simulation In Computer Games

Abstract

Multi Agent System (MAS) have been in existence for quite a long time and have been a focused area of research in different paradigms over the years. As a result, several test beds and simulation tools exist for their deployment in software engineering problems. However, these test beds are suited for specific agents types and environments and lack support for interoperating with other agent architectures and environments. Further, they do not exploit the power of modern distributed and parallel computing environments. Here at RIATec, there is a powerful escience Beowulf cluster that provides ample resources for distributed simulation of MAS. Such simulation has several possible target applications such as in computer games and training virtual environments and therefore it is important that such simulation is reused. Moreover, good visualization, rendering and interactive tools are needed for evaluating, testing and interacting with the system. In this paper, a proposal for a reusable architecture and visualization test bed based on Highlevel architecture (HLA) and Unreal technology is provided for a high performance MAS simulator.