论文标题
部分可观测时空混沌系统的无模型预测
Digital twins for city simulation: Automatic, efficient, and robust mesh generation for large-scale city modeling and simulation
论文作者
论文摘要
储层计算是预测湍流的有力工具，其简单的架构具有处理大型系统的计算效率。然而，其实现通常需要完整的状态向量测量和系统非线性知识。我们使用非线性投影函数将系统测量扩展到高维空间，然后将其输入到储层中以获得预测。我们展示了这种储层计算网络在时空混沌系统上的应用，该系统模拟了湍流的若干特征。我们表明，使用径向基函数作为非线性投影器，即使只有部分观测并且不知道控制方程，也能稳健地捕捉复杂的系统非线性。最后，我们表明，当测量稀疏、不完整且带有噪声，甚至控制方程变得不准确时，我们的网络仍然可以产生相当准确的预测，从而为实际湍流系统的无模型预测铺平了道路。
The concept of creating digital twins, connected digital models of physical systems, is gaining increasing attention for modeling and simulation of whole cities. The basis for building a digital twin of a city is the generation of a 3D city model, often represented as a mesh. Creating and updating such models is a tedious process that requires manual work and considerable effort, especially in the modeling of building geometries. In the current paper, we present a novel algorithm and implementation for automatic, efficient, and robust mesh generation for large-scale city modeling and simulation. The algorithm relies on standard, publicly available data, in particular 2D cadastral maps (building footprints) and 3D point clouds obtained from aerial scanning. The algorithm generates LoD1.2 city models in the form of both triangular surface meshes, suitable for visualisation, and high-quality tetrahedral volume meshes, suitable for simulation. Our tests demonstrate good performance and scaling and indicate good avenues for further optimization based on parallelisation. The long-term goal is a generic digital twin of cities volume mesh generator that provides (nearly) real-time mesh manipulation in LoD2.x.
