DDDAS-SEP: Dynamically-Integrated Production Planning and Operational Control for the Distributed Enterprise

现代社会依赖于许多相互作用的大规模动态系统（系统的系统），其中大多数太复杂，无法进行数学分析。这种系统网络的行为取决于它们之间的联系和环境，即科学还原论无法完全定义行为。这些应用的典型特征是分层控制和水平的、相互依赖的组件（联邦），并且在动态环境中要求整个或子系统以准确和及时的方式自适应地响应变化以提高性能。设想的总体方案是一个多尺度的交织的仿真和决策模型，支持规划和控制，通过感测和流数据的动态更新。规划和控制模型直接从真实世界的传感数据和模拟中更新。虽然真实的世界可能会产生诸如当前成本或库存状态之类的参数，但与代表供应网络中其他实体的其他联邦同步的仿真模型将用于计算更复杂的模型参数，诸如前置时间和有效资源可用性。此外，仿真模型将被离线使用，以测试各种建议的决策规则/模型的性能，并指导在每个重新规划实例的模型构建和选择的选择。重新规划将发生时，实际观察和模型形式之间的不一致性达到一个定义的阈值，用于构建主动planne.To实现这些目标的项目将研究方法的发展，为分布式计算在网络传感器驱动，分层决策和控制系统的分布式企业的集成架构。这种架构，这是基于无处不在的Web服务技术，将允许集成的多尺度模型，支持模拟，决策和控制功能，与传感器进行交互。建议的架构将促进集成在几个方面：开发一个连续的传感器驱动的方法来规划和控制协调;互操作性的分布式系统的不同类型的模型;集成的传感器/执行器与各种模型的相互作用;集成不同类型的传感器/执行器，符合不同的标准;集成的传感器测量模型验证过程。

Modern society depends upon many interacting large-scale dynamic systems (systems of systems), most of which are too complex for mathematical analysis. The behavior of such system networks depends on their linkages and the environment, i.e. scientific reductionism is incapable of fully defining behavior. These applications are typically characterized by hierarchical control and horizontally, interdependent components (federates), and in a dynamic environment require the whole or sub system(s) to respond adaptively to changes in an accurate and timely manner for improved performance.This project focuses on hierarchical production and logistics planning and control in highly capitalized discrete manufacturing system networks. The overall scheme envisioned is a multi-scale federation of interwoven simulation and decision models that support planning and control with dynamic updating through sensed and streamed data. Planning and control models are updated both directly from real-world sensed data and from simulations. While the real world may yield parameters such as current costs or inventory status, the simulation model, in synchronization with other federates representing the other entities in the supply network, will be used to compute more complex model parameters such as lead times and effective resource availabilities. In addition, the simulation model will be used off-line to test the performance of various proposed decision rules/models and guide the selection of model construction and selection at each re-planning instance. Re-planning will occur when a defined threshold of discordance between actual observation and model form used to construct the active plan is reached.To accomplish these goals the project will investigate methods for the development of an integrated architecture for distributed computing in networked sensor-driven, hierarchical decision and control systems of the distributed enterprise. This architecture, which is based on ubiquitous web service technology, will allow an integrated collection of multi-scale models that support the simulation, decision and control functions to interact with transducers. The proposed architecture will facilitate integration in several ways: developing a continuous sensor-driven approach to planning and control coordination; interoperability of different genres of models for distributed systems; integration of sensor/actuator interactions with a variety of models; integration of different genres of sensors/actuators that conform to different standards; integration of sensor measurements in the model validation process.