Gal Barkai will speak at the Control and Signal Processing (CSP) seminar at the University of Melbourne. His talk is about his recent work exploring new architectures for consensus protocols.
Title: Beyond Classical Consensus: Sampled Information Exchange and Two Degrees-of-Freedom Approaches
Abstract: Over the past two decades, considerable research in systems and control has focused on networked multi-agent systems, where the central challenge lies in designing controllers under communication constraints such as transmission delays, sampled communication, and sparse information exchange topologies. At the core of this field is the agreement problem and its celebrated solution—the consensus protocol. By directly incorporating the information exchange graph into the controller, the consensus protocol enables the use of powerful tools from algebraic graph theory for system analysis. As a result, it has become the de facto control structure for agreement problems, even in highly complex settings. In this talk, we argue that despite their elegance and simplicity, consensus-based controllers can be unnecessarily conservative. To illustrate this point, we examine two scenarios: (i) when information exchange is sampled, and (ii) when transmitted information is corrupted by noise. Drawing intuition from classical areas such as optimal sampled-data control and servo regulation, we propose modified control structures that simplify both the analysis and the design process. This talk is based on joint work with Leonid Mirkin and Daniel Zelazo from the Technion.