Hybrid Models for Simulation-Based Prototyping of Decision Logic in Material Flow Systems

这笔赠款提供资金，以创建一种新的离散事件模拟方法，这是一种广泛用于工厂和仓库等离散物质流系统的设计和运营改进的计算工具。这一新方法结合了现有的两种模式。第一种是流程交互范例，它有效地对系统中的物料流、客户流或工作流进行建模，但在表示复杂的决策逻辑方面存在局限性。第二种是事件调度范例，该范例用于对复杂系统进行建模，但在建模时间和精力方面非常昂贵。通过对离散物流系统的物理和决策逻辑特征进行分类，为不同类别的系统指定通用的决策逻辑单元，使用事件调度来封装决策行为。这些单元将与物理系统流的流程交互表示相连接。这项研究将在实验上将使用新的混合方法获得的结果与使用更简单模型获得的结果在系统性能估计方面进行比较。由于计算时间是模拟的主要关注点，该项目还将测量与更简单的系统表示相比，拥有这样的决策逻辑单元的计算效果。该项目的成功完成以及由此产生的模拟建模方法的混合方法将提高具有复杂决策逻辑的系统建模和分析的技术水平。一个主要目标是提供更好的原型能力，在决策逻辑在当今日益复杂的系统中实现之前对其进行测试。通过将混合方法与跨各种系统类型的更简单的表示进行比较，可以初步了解建模中细节级别的影响。其他预期结果包括对研究生和本科教育的影响，以及向研究界提供的案例研究。

This grant provides funding to create a new approach to discrete-event simulation, a computational tool used extensively for design and operational improvement of discrete material flow systems such as factories and warehouses. This new approach combines two existing paradigms. The first is the process-interaction paradigm, which efficiently models flow of material, customers or work through a system, but has limitations in representing complex decision logic. The second is the event-scheduling paradigm, which is used to model complex systems, but is quite costly in modeling time and effort. By classifying discrete material flow systems according to physical and decision logic characteristics, generic decision logic units will be specified for different classes of systems, using event-scheduling to encapsulate decision-making behavior. These units will be interfaced with a process-interaction representation of the physical system flow. The research will experimentally compare results obtained using the new hybrid approach against those obtained using simpler models, in terms of estimates of system performance. Since computation time is a major concern with simulation, the project also will measure the computational effect of having such decision logic units, as compared to a simpler system representation.Successfully completed, this project and the resulting hybrid approach to simulation modeling methodology will improve the state of the art in modeling and analyzing systems with complex decision logic. One major goal is to provide a better prototyping capability, to test decision logic before it is implemented in today's increasingly complex systems. By comparing the hybrid approach against simpler representations across a variety of system types, an initial understanding of the effect of level of detail in modeling will be achieved. Additional expected results include impact on graduate and undergraduate education, and case studies made available to the research community.