Resilient Function Calculation in Distributed Control Systems

Abstract

The talk addresses aspects of the problem of reliable/resilient coordination and control of distributed systems over shared infrastructures. This type of systems have sustained tremendous growth rates recently due to the proliferation of digital and networking technologies, and include a variety of sensor/actuator networks, such as electric power distribution systems (e.g., smart grids), robotic networks, and transportation systems of various sorts. More specifically, we consider a distributed system whose components (nodes) can locally exchange information in an iterative fashion via interconnections (edges) that form an arbitrary, possibly directed topology (digraph). We focus on linear iterative updates in which the nodes maintain and update certain values based on linear combinations of their own values and the values they successfully receive from their neighbors. By employing appropriate encoding of the data and/or by taking advantage of redundant paths in the topology of the network, we develop resilient iterative strategies to compute functions (such as the average) of the initial values that the nodes posses, despite unreliable (packet-dropping) links and/or incorrect actions by faulty/malicious nodes. These calculations can be used as primitives for providing resilience to a variety of distributed control, optimization, and coordination tasks.