Networked control systems are ubiquitous in modern applications, ranging from self-driving cars, autonomous supply-and-delivery systems, to spacecraft formation flight. In this talk, we address algorithms for controlling such systems, and tools from algebraic graph theory that allow us to optimize the performance of those algorithms. We examine how these algorithms reject external noise signals in the context of the H2 system norm. First, a fast algorithm is discussed for computing the H2 norm of consensus on series-parallel networks. Extensions to multiplex, matrix-weighted graphs are presented. For networks with non-heterogenous agents with system dynamics operating at different timescales, we examine how the graph topology affects the H2 norm of edge consensus, a representation of the consensus algorithm using relative measurements between agents. We present algorithms that find the minimum-H2 norm spanning tree of a graph, and discuss how notions of graph centrality affect the H2 norm when choosing to add additional connections between nodes.
- This event has passed.
18 September Friday
Mathias Hudoba de Badyn (ETH): Optimizing Graphs for Networked Dynamical Systems
18 September 2020, 11:00 - 18 September 2020, 12:00
En ligne/online, Canada