Scalable Enterprise Systems: Scalable Algorithms and Distributed Agent Architectures for Adaptive Enterprises

该拨款为在Internet上使用分布式代理体系结构的新型自适应制造企业构建高度可伸缩的分布式算法提供资金。这些算法将用于重新配置在制品库存水平和生产计划，以适应不断变化的市场需求和供应链条件。此外，这些算法和必要的信息将嵌入到分布在整个企业的地理位置上的代理中。目标是保持响应性和有效性，以支持下一代可伸缩企业。可扩展性将通过(1)可预测且计算效率高的分布式算法实现；(2)支持规模和能力增长的分布式代理架构；(3)分布式集群，使用成本效益高的技术提供信息的快速访问。将开发分析模型来预测此类系统的紧急行为及其计算的稳定性和收敛性。这些模型将用于评估计算复杂性、通信需求和算法的可扩展性。分布式代理体系结构将基于任务分解和任务相似度来识别相似的代理集群。使用商用pc集群的实验原型将用于基准计算负载，通信网络需求和算法的整体可扩展性。如果成功，本研究的结果将导致车间与企业软件的有效集成，这是有效和优化管理多个企业的关键问题。这些研究活动将涉及与几个行业伙伴的积极合作。在短期内，这项工作将提供一套新颖的分布式算法和合适的分布式代理体系结构。从长远来看，这项工作将通过建立对可扩展性的科学和技术见解，为设计下一代自适应和可重构企业提供数学和仿真模型。

This grant provides funding for building highly scalable distributed algorithms for a new class of adaptive manufacturing enterprises using distributed agent architecture over the Internet. These algorithms will be used for reconfiguring work-in-process inventory levels and production schedules to adapt to changing market demands and supply-chain conditions. Moreover, these algorithms and the necessary information will be embedded in agents geographically distributed throughout the enterprise. The goal is to maintain responsiveness and effectiveness to enable the next generation of scalable enterprises. Scalability will be achieved through (1) distributed algorithms that are predictable and computationally efficient; (2) distributed agent architectures that support growth in size and capability; and (3) distributed clusters that provide rapid access to information using cost effective technologies. Analytical models will be developed to predict the emergent behavior of such systems and stability and convergence properties of their computations. These models will be used for assessing computational complexity, communication requirements, and scalability of the algorithms. Distributed agent architecture will be developed in which clusters of similar agents will be identified based on task decomposition and task similarity. An experimental prototype using a cluster of commodity PCs will be used to benchmark the computational load, communication network requirements, and overall scalability of the algorithms.If successful, the results of this research will lead to effective integration of shop floor with enterprise software which is a critical issue in managing multiple enterprises effectively and optimally. These research activities will involve active collaboration with several industry partners. In the short-term, this work will provide a set of novel distributed algorithms and a suitable distributed agent architecture. In the long-term, this work would provide mathematical and simulation models for designing the next generation of adaptive and reconfigurable enterprises by establishing scientific and technological insights into their scalability.