 |
Center for Nonlinear Studies |
This talk focuses on distributed control of dynamical flow networks. These are modeled as dynamical systems derived from mass conservation laws on directed capacitated networks. The flow evolution through the network is governed by routing, scheduling, and flow control policies within constraints imposed by the network infrastructure and physical laws. Depending on the application (e.g., data networks, road traffic networks, distribution networks), such policies are meant to represent local controls, users’ behavior, or a combination of the two. Some versions of these models include cascading failures mechanisms, whereby overloaded links become inactive and potentially cause the overload and failure of other nodes and links in the network. We focus on efficiency, resilience, and scalability properties of such dynamical flow networks. First, we show that optimal throughput and resilience can be achieved by feedback policies that depend only on local information and require no global knowledge of the network. Then, we prove how the optimal selection of a stable equilibrium and the optimal control of the transient can be cast as convex problems which are amenable to distributed solutions. Applications to arterial traffic control are discussed.
Host: Misha Chertkov