Decision making in smart networks
Tid: Må 2017-05-22 kl 11.00 - 12.00
Föreläsare: Dina Irofti
Plats: Automatic control Dep. The Netcon seminar room
Abstract: This talk presents a new consensus algorithm for networks with smart agents. Every agent in the network can anticipate its neighbours' opinion based on past information (past states) and adapts its own opinion in consequence. The condition under which the network can reach consensus is studied. The results show that this condition is strongly linked to how the agents use the past information, and that, by making the right choice, this anticipatory algorithm can improve the speed of consensus when compared to classical consensus problem, especially in networks with poor connectivity. Moreover, it turns out that if in the case of undirected networks the condition for consensus reaching does not depend on the network's type, for directed networks this condition depends on the topology in a nontrivial way.
Bio: Dina Irofti received her B.Sc. degree in Automatic Control and Computers from Polytechnic Institute of Bucharest, Romania in 2013 and her M.Sc. degree in Automatic Control from Supélec (now known as CentraleSupélec), France in 2014. Since 2014 she is a PhD candidate in the Laboratory of signals and systems, France, working on delays and networked systems. During her PhD, she received a DAAD grant and worked in 2015 as a research assistant at Max Planck Institute for Mathematics, Germany, studying consensus problem; in 2016 she received a Mitacs Globalink grant for a fellowship at the University of Lethbridge, Canada, carrying interdisciplinary research on biochemical models.