Large-Scale Computing in Control

控制中的大规模计算

• 批准号: 9120643
• 负责人: Alan Laub
• 金额: $ 21万
• 依托单位: University of California-Santa Barbara
• 依托单位国家: 美国
• 项目类别: Continuing grant
• 财政年份: 1992
• 资助国家: 美国
• 起止时间: 1992-08-01 至 1996-07-31
• 项目状态: 已结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=9120643&HistoricalAwards=false
• 关键词: Large; Scale; Computing; Control

In this proposal we discuss algorithms for solving high-order numerical problems that arise in the design of control laws for systems modeled with large numbers of variables (parameters, inputs, outputs, states, etc.). Particular attention is focused on certain key matrix equations arising in state- space-based control engineering. The most common such equations are Lyapunov and Riccati equations, and their effective numerical solution is described in detail, although our techniques also apply more generally. Considerable discussion is focused on large-scale matrix problems, including those in which the matrices are sparse or structured. In addition to continuing our research on algorithms for implementation on vector supercomputers such as the Cray Y-MP for which coarse-grained parallelism can also be exploited, we proposing now to extend our efforts to computers such as the Connection Machine with massive parallelism. Much of our proposed research is intended to lay the groundwork for the next generation of computer-aided control system design and analysis software.

在这个方案中，我们讨论了在设计具有大量变量(参数、输入、输出、状态等)的系统的控制律设计中出现的高阶数值问题的算法。特别关注基于状态空间的控制工程中出现的某些关键矩阵方程。最常见的此类方程是Lyapunov方程和Riccati方程，并详细描述了它们的有效数值解，尽管我们的技术也适用于更广泛的情况。大量的讨论集中在大规模的矩阵问题上，包括其中的矩阵是稀疏的或结构化的。除了继续研究在诸如Cray Y-MP之类的向量超级计算机上实现的算法之外，还可以利用粗粒度并行性，我们现在提议将我们的努力扩展到具有大规模并行性的连接机等计算机。我们提出的许多研究都是为了为下一代计算机辅助控制系统设计和分析软件奠定基础。