The talk will describe three different classes of linear network
models that we have been studying, particularly to understand how
network structure affects dynamics, control and estimation. The first of these model classes represents the so-called swing dynamics of a power network, but also shows the possibility of wave (rather than swing) motions, for which one can develop zero-reflection controllers. The second model class represents a network of Markov chains whose transitions are influenced in a particular way by the current states of their neighbors. The model can display cascading phenomena. Some generalizations of hidden Markov models, and related estimation problems, also emerge. The third model class represents a traffic network whose flows are controlled by stochastically specified controllers. Worthwhile comparisons can be made with fluid models of
queueing networks.
Copyright. All Rights Reserved.