SI2-SSE: Software Elements to Enable Immersive Simulation

并行计算机已经变得如此强大，它们现在能够在几秒钟内解决极其复杂的流体流动或结构问题。不幸的是，在解决一个复杂的问题之前，研究人员可能需要花费许多小时甚至几天的时间来建立一个复杂的问题。此外，如果使用标准技术，可能需要数小时或数周才能从模拟产生的大量数据中提取洞察力。对于发现和设计问题，问题的下一个变体需要更改问题定义，这些延迟破坏了实验流程和相关的直觉，以及了解问题定义的更改如何与解决方案的更改相关。为了解决这个问题，一种范式转变，在这里被称为“沉浸式仿真”，计划启用问题定义编辑的新方法，允许从业者以一种方式与仿真（可视化模型迭代）交互，在这种方式下，他们可以动态地体验单个、实时和正在进行的仿真参数变化的影响。例如，外科医生在一台电脑显示器上虚拟地改变了旁路移植的形状，然后虚拟地观察到不仅在旁路内，而且在整个血管系统中血流模式的变化。同样，工程师改变虚拟汽车的形状，看看流动模式是改善还是恶化。这些应用研究实例在基础研究中有相似之处，在基础研究中，研究人员将首次获得对非定常、湍流流动物理及其对动态参数变化的敏感性的实时洞察。可视化地将解决方案的变化与视觉迭代的几何和/或参数变化联系起来，将开创一个由直觉驱动的发现和设计的新时代。这种模式的转变也将被纳入基础本科和研究生课程，以实现更深入的、基于体验的学习。该项目的中心目标是将最先进的工具推进到通用组件中，当集成时，将为任何偏微分方程求解器提供以下功能：1)实时的、可重构的正在进行的仿真可视化；2)实时的、可重构的问题定义，允许动态解洞察力来指导关键问题参数的选择；3)实时参数灵敏度反馈；4)自适应仿真控制，以考虑离散化误差和几何变化；5)可靠的、沉浸式仿真的集成和演示。这些软件组件开发的第一个社区包括心血管流和空气动力学流控制。他们已经明确表示需要软件在具有直观和定量参数敏感性的广泛参数空间下更快速地探索其系统的性能。该软件不仅可以进行设计（应用研究，例如，探索旁路与支架类型和特定患者病变血管的放置或流量控制执行器的放置），还可以进行发现（基础研究，例如，探索流量响应的物理学，以发现全新的外科手术程序和流量控制过程和设备）。这种双重和互补的软件应用程序将对教育产生类似的影响，基础课程将使用集成的软件模块来创建沉浸式模拟，以建立对流物理的直觉，然后在顶点设计课程中加强应用性质的学习。虽然这些想法将在流体动力学领域进行原型化和验证，但它们将通过可持续的软件工程进行普遍开发，以便于其他使用模拟的领域采用。这些工具的成功开发、集成和大规模演示将把大规模并行仿真从一系列I/ o密集型步骤转变为使用精心设计的接口进行实时、可重构的发现，为所有仿真开辟了道路。该奖项反映了美国国家科学基金会的法定使命，并通过使用基金会的知识价值和更广泛的影响审查标准进行评估，被认为值得支持。

Parallel computers have grown so powerful that they are now able to solve extremely complex fluid flow or structures problems in seconds. Unfortunately, it may take a researcher many hours or even days to set up a complex problem before it can be solved. Furthermore it may take hours or often weeks to extract insight from the volume of data the simulation produces, if using standard techniques. For discovery and design questions, where the next variant of the problem requires a change to the problem definition, these delays disrupt the flow of experimentation and the associated intuition and learning about how the change in the problem definition relates to a change in the solution. To address this issue, a paradigm shift, referred to here as "immersive simulation", is planned to enable new approaches to problem definition editing that allow practitioners to interact with the simulations (visual model iteration) in a manner where they can dynamically experience the influence of parameter variations from a single, live, and ongoing simulation. Examples include a surgeon virtually altering the shape of a bypass graft on one computer monitor and then virtually observing the change in the blood flow patterns not only within the bypass but throughout the vascular system. Likewise, an engineer altering the shape of a virtual car to see if the flow pattern improves or worsens. These applied research examples have parallels in fundamental research where live insight into the flow physics of unsteady, turbulent flows and their sensitivity to live parameter changes will be made available to researchers for the first time. Visually connecting the solution change to the visually iterated geometry and/or parameter change will enable a new age of intuition-driven discovery and design. This paradigm shift will also be incorporated into foundational undergraduate and graduate courses to enable deeper, experiential-based learning. The central goal of this project is to advance state-of-the-art tools into generic components that, when integrated, will make the following capabilities available to any partial differential equation solver: 1) live, reconfigurable visualization of ongoing simulations, 2) live, reconfigurable problem definition to allow the dynamic solution insight to guide the choice of key problem parameters, 3) real-time parameter sensitivity feedback, 4) adaptive simulation control to account for discretization errors and geometry changes, and 5) integration and demonstration of reliable, immersive simulation. The first communities that these software components will be developed with include cardiovascular flow and aerodynamic flow control. They have already articulated a need for software to more rapidly explore the performance of their systems under a broad parameter space with intuitive and quantitative parameter sensitivity. This software will enable not only design (applied research e.g., exploring bypass vs. stent type and placement for a particular patient's diseased vasculature or flow control actuator placement), but also discovery (fundamental research e.g., explore physics of flow response to discover completely new surgical procedures and flow control processes and devices). This twofold and complementary software application will have a similar impact on education, where foundational courses will use the integrated software modules to create immersive simulations that build intuition about flow physics, and then reinforce that learning in an applied nature in capstone design courses. While the ideas will be prototyped and proven within the field of fluid dynamics, they will be developed generally, with sustainable software engineering, for easy adoption by other fields that make use of simulation. The successful development, integration, and demonstration of these tools at scale will transform massively parallel simulation from a series of I/O-intensive steps to live, reconfigurable discovery using carefully designed interfaces that blaze the trail for all simulation.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.