Scalable Enterprise Systems Phase II: Emergent Organizational Structures in Multi-Agent Scalable Enterprises

这个可扩展企业系统第二阶段项目的主要目标是开发新兴企业模型，捕获所涉​​及组织的独立行为以及各个组织之间交互的影响，以便准确预测新兴企业的动态和系统的长期性能。 企业参与者可以使用这些结果来选择适当的运营策略，以改进本地绩效指标，例如固定规划期限内的利润或现金流。 此外，这些模型将提供对导致有效联盟和组织设计的机制的深入了解。 建模方法分为三个步骤。 首先，开发了一种表示方案，用于捕获个体参与者的特征，其中包含任务信息、组织关系、本地和系统级目标以及环境中可能的变化。 因此，企业是每个组织以及这些组织之间的交互的多个此类详细表示的集合。 其次，开发了一种微观-宏观建模方法，其中每个代理的任务细节（在微观模拟框架中表示）被聚合成宏观层面的一组交互参数。宏观层面的企业模型被表述为一个相互作用的粒子系统。最后，第三步涉及验证使用现场数据开发的理论。将从供应网络中有大量参与者的不同行业（例如食品行业和汽车行业）收集现场数据，以构建实际的代理表示并验证理论性能估计。使用多代理模拟器的大规模模拟也将用于凭经验估计给定组织配置的性能。这项研究将使规划者能够先验地估计不同运营条件下的整体企业绩效，并最终设计最佳的新兴社会。此外，由分布式决策者组成的系统在各种应用领域中都会遇到，例如高速公路网络上的交通、空中交通管理系统、由分散设计者组成的设计团队、分布式传感器网络、多智能体系统等。这项研究取得的根本进展将对我们预测此类系统性能的能力产生重大影响，从而制定适当的规划和控制策略。

This Scalable Enterprise Systems Phase II project's primary goal is to develop models of emergent enterprises that capture the independent behavior of the organizations involved, as well as the effects of interactions among individual organizations, in order to accurately predict dynamics of the emergent enterprise and performance of the system in the long run. Participants in the enterprise can use these results to select appropriate operational strategies in order to improve local performance measures such as profit or cash flow over a fixed planning horizon. In addition, these models will provide insight into the mechanisms that result in effective alliances and organizational designs. The modeling approach has three steps. First, a representation scheme is developed for capturing characteristics of individual participants that incorporates task information, organizational relationships, local and system level goals, and possible changes in the environment. An enterprise is thus a collection of several of such detailed representations for each organization and the interactions between these organizations. Second, a micro-macro modeling approach is developed, where the details of each agent's tasks (represented in the micro-level simulation framework) are aggregated into a set of interaction parameters at the macro-level. The macro-level enterprise model is formulated as an interacting particle system. Finally, the third step involves validation of the theories developed using field data. Field data will be collected from different industries that have large number of participants in the supply network, such as the food industry and the automotive industry, to construct practical agent representations and to verify the theoretical performance estimates. Large-scale simulations using a multi-agent simulator will also be used to empirically estimate the performance of a given organizational configuration. This research will allow planners to estimate a priori the overall enterprise performance under different operating conditions, and eventually, design optimal emergent societies. Moreover, systems composed of distributed decision-makers are encountered in various application domains, such as traffic on highway networks, air traffic management systems, design teams composed of dispersed designers, distributed sensor networks, multi-agent systems, etc. The fundamental advances made in this research will have significant impact on our ability to predict the performance of such systems, and therefore devise appropriate planning and control strategies.