A High-Performance Open Testbed for Multi-Agent Systems

该项目将为大规模、大粒度多智能体软件系统（MAS）的实验研究创建一个可共享的科学基础设施。该研究将扩展现有的轻量级，灵活，集成和可扩展的实验平台MACE 3 J的功能。在当前原型的基础上，将在MACE 3 J中设计额外的功能，并添加几个建模/分析组件库。 由此产生的基础设施将进行测试的鲁棒性和可扩展性的小到大规模的多计算机平台上使用一套示例MAS的问题。研究产生的工具和文件将作为社区研究基础广泛分发。 成果将包括：2）一种技术，它将开始把MAS社区的努力与正在进行的高性能计算研究统一起来，这些研究领域包括单系统映像集群、非常大的分布式计算网格、高度异构的高性能代码以及使用Java进行高性能计算; 3）多智能体系统研究和结果的教育和演示工具; 4）共同基准问题的基础，以提供结果的可比性，并帮助衡量MAS领域的总体进展。

This project will create a sharable scientific infrastructure for experimental research on large-scale, large-grain multi-agent software systems (MAS). The research will extend the capabilities of an existing lightweight, flexible, integrative, and scalable experimental platform called MACE3J. Building on the current prototype, additional functionality will be designed into MACE3J, and several modeling/analysis component libraries will be added. The resulting infrastructure will be tested for robustness and scalability on small-to-large-scale multicomputer platforms using a suite of example MAS problems. The tools and documentation produced by the research will be widely distributed as a community research base. Outcomes will include: 1) a facility for collaboration and model/data integration within the MAS research community; 2) a technology that will begin unifying MAS community efforts with ongoing high-performance computing research in areas such as single-system-image clustering, very large, distributed computational grids, highly heterogeneous high-performance codes, and the use of Java for high-performance computing; 3) tools for education and demonstration of multi-agent systems research and results; and 4) foundations for common benchmark problems, to provide comparability of results and to help measure general progress in the field of MAS.