Collaborative Research: Elements: Software: NSCI: Chrono-An open-source simulation platform for computational dynamics problems

该项目旨在增强Chrono所采用的建模和解决方法，Chrono是一个多体动力学（MBD）和流固相互作用（FSI）问题的开源计算机仿真平台。Chrono将能够捕捉从毫秒（撞击现象）到几十年（地球物理）不等的各种尺寸和时间尺度的动态。这些性能水平在几个领域开辟了新的研究方向。Chrono被其他用户广泛使用和进一步开发，目前有一个活跃的论坛，拥有250多名注册用户。该项目将增强Chrono建模功能的丰富程度，完善的数值解决方案基础，并利用新兴的硬件架构，将这种模拟能力提升到即用型、开源、一流的计算动力学平台的地位。Chrono已被大学、国家实验室和工业使用。在过去的两年中，不同的团队已经将Chrono应用于地外应用、机器人机器学习、图像处理、模式识别和计算机视觉、机械表设计、建筑研究、自动驾驶汽车、流固交互应用、风力涡轮机动力学、下一代太空服设计、石油开采和事故缓解、硬件在环仿真等领域。最后，该项目将吸引来自弱势群体的高中生参加为期六天的住宿夏令营（现在是第12届），并将通过一个新的住宿项目培训一组来自威斯康星大学麦迪逊分校的加州州立大学的本科生，该项目将向他们介绍在基于仿真的机器人设计中使用Chrono。该项目旨在增强Chrono采用的建模和解决方法，Chrono是BSD3开源仿真平台，用于多体动力学（MBD）和流固相互作用（FSI）问题。该项目的软件基础设施增强旨在维持具有超过一百亿个自由度的MBD和FSI系统的万亿次浮点级仿真；也就是说，比今天的传统模拟高出两到三个数量级。为了提高采用率和影响，目标性能水平将在利用GPU计算的预算/负担得起的硬件上达到。Chrono将能够在从毫秒（撞击现象）到几十年（地球物理学）的时间尺度上捕捉微观、中观和宏观尺度的动态。本项目的智力价值源于以下关键思想：(i)着眼于摩尔定律的日落，软件设计解决方案包含可扩展的多gpu硬件布局，可以有效地解决大型多物理场问题；（ii）硬件感知软件设计范例，它积极减少数据存储和移动，将允许预算敏感的硬件系统运行十亿自由度模型，或者，对于类似大小的模型，与最先进的模型相比，实现两个数量级的加速；（三）固体和流体动力学的统一拉格朗日公式在一个软件平台上实现，可以模拟复杂的多物理场（耦合）问题；（iv） Chrono提出了一种处理摩擦和接触的替代方法，该方法围绕微分变分不等式的概念展开，从而避免了阻碍大多数现有多体动力学模拟器的小积分时间步长和数值不稳定性问题。就其教育和推广举措而言，该项目：(a)将有助于建立威斯康星大学麦迪逊分校的一门新的本科课程，向学生介绍计算机概念，随后在研究生高级计算机课程中加以完善；(b)将通过每年两次面向代表性不足的学生的暑期住宿项目，在高中和本科阶段促进计算科学和计算动力学学科的发展；(c)将扩大一个先进计算论坛，以促进向工业界转让技术和促进Chrono的采用；(d)将加强目前在机器人、地质力学和软物质物理学方面严重依赖Chrono的合作。目前，Chrono平均每天被克隆10次，有150多个参与方从其公共存储库进行分叉，并拥有一个拥有250多名注册用户的活跃论坛。该项目将增强Chrono建模功能的丰富程度，改进其数值解决方案基础，并利用新兴的硬件架构将其仿真能力提升到即用型、开源、一流的计算动力学平台的地位。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。

This project seeks to augment modeling and solution methods employed by Chrono, an open-source computer simulation platform for multi-body dynamics (MBD) and fluid-solid interaction (FSI) problems. Chrono will be able to capture dynamics at various size and time scales spanning from millisecond (impact phenomena) to decades (geophysics). These performance levels open up new directions of research in several fields. Chrono is widely used and further developed by other users and has an active forum with more than 250 registered users currently. This project will enhance the richness of Chrono's modeling features, sound numerical solution foundation, and leverage of emerging hardware architectures to elevate this simulation capability to the status of ready-to-use, open-source, best-in-class computational dynamics platform. Chrono has been used by universities, national labs, and industry. Over the past two years, various groups have used Chrono in extraterrestrial applications, machine learning in robotics, image processing, pattern recognition and computer vision, mechanical watch design, architectural studies, autonomous vehicles, fluid-solid interaction applications, wind turbine dynamics, next generation space suit design, oil extraction and accident mitigation, hardware-in-the-loop simulation, etc. Finally, this project will engage high-school students from under-represented groups in a six-day residential camp run (now at its 12th edition) and will train a group of undergraduate students from California State University at University of Wisconsin-Madison through a new residential program that will introduce them to the use of Chrono in simulation-based robotics design.This project seeks to augment modeling and solution methods employed by Chrono, a BSD3 open-source simulation platform for multi-body dynamics (MBD) and fluid-solid interaction (FSI) problems. The software infrastructure enhancements in this project aim at sustaining teraflops-grade simulation of MBD and FSI systems with more than ten billion degrees of freedom; i.e., two to three orders of magnitude beyond conventional simulations today. In order to increase adoption and impact, the performance levels aimed at will be reached on budget/affordable hardware that leverages GPU computing. Chrono will be able to capture micro-, meso- and macro-scale dynamics on time scales spanning from millisecond (impact phenomena) to decades (geophysics). The intellectual merit of this project stems from the following key ideas: (i) with an eye towards the sunsetting of Moore's law, the software design solution embraces a scalable multi-GPU hardware layout poised to solve effectively large multi-physics problems; (ii) a hardware-aware software design paradigm, which aggressively reduces data storage and movement, will allow budget-conscious hardware systems to run billion-degree-of-freedom models, or, for models of similar size, accomplish a two orders of magnitude speedup when compared to the state of the art; (iii) a unified Lagrangian formulation for both solid and fluid dynamics is implemented in one software platform that can simulate complex multi-physics (coupled) problems; and (iv) Chrono promotes an alternative approach for handling friction and contact that revolves around the concept of differential variational inequality and thus avoids the small integration time step and numerical instability issues that hinder most of the existing many-body dynamics simulators. In relation to its educational and outreach initiatives, this project: (a) will be instrumental in establishing a new University of Wisconsin-Madison undergraduate course that introduces students to computing concepts subsequently refined in a graduate advanced computing class; (b) will promote the discipline of Computational Science and Computational Dynamics at high-school and undergraduate levels via two yearly residential summer programs for under-represented students; (c) will expand an advanced computing forum that facilitates technology transfer to industry and promotes Chrono adoption; and, (d) will strengthen ongoing collaborations that critically depend on Chrono in robotics, geomechanics, and soft-matter physics. Chrono is presently cloned on average 10 times every day, has been forked from its public repository by more than 150 parties, and has an active forum with more than 250 registered users. This project will enhance the richness of Chrono modeling features, improve its numerical solution foundation, and leverage emerging hardware architectures to elevate this simulation capability to the status of ready-to-use, open-source, best-in-class computational dynamics platform.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.