Licentiate Defense at Automatic Control: Jonas Hansson

Date & Time: September 8th, 10:15-12:00
Location: Room M:Ångpannehallen (M:1328), in the M-building, LTH
Title: Transient analysis and control for scalable network systems
Opponent: Bart Besselink, University of Groningen
Examiner: Yiannis Karayiannidis
Zoom link: https://lu-se.zoom.us/s/68971654803
Thesis PDF: LU Publications

Abstract: The rapidly evolving domain of network systems poses complex challenges, especially when considering scalability and transient behaviors. This thesis aims to address these challenges by offering insights into the transient analysis and control design tailored for large-scale network systems. The thesis consists of three papers, each of which contributes to the overarching goal of this work.

The first paper, A closed-loop design for scalable high-order consensus, studies the coordination of n:th-order integrators in a networked setting. The paper introduces a novel closed-loop dynamic named serial consensus, which is designed to achieve consensus in a scalable manner and is shown to be implementable through localized relative feedback. In the paper, it is shown that the serial consensus system will be stable under a mild condition -- that the underlying network contains a spanning tree -- thereby mitigating a previously known scale fragility. Robustness against both model and feedback uncertainties is also discussed.

The second paper, Closed-loop design for scalable performance of vehicular formations, expands on the theory on the serial consensus system for the special case when n=2, which is of special interest in the context of vehicular formations. Here, it is shown that the serial consensus system can also be used to give guarantees on the worst-case transient behavior of the closed-loop system. The potential of achieving string stability through the use of serial consensus is explored.

The third paper, Input-output pseudospectral bounds for transient analysis of networked and high-order systems, presents a novel approach to transient analysis of networked systems. Bounds on the matrix exponential, coming from the theory on pseudospectra, are adapted to an input-output setting. The results are shown to be useful for high-order matrix differential equations, offering a new perspective on the transient behavior of high-order networked systems.