Friday, October 28 at 11:00 am in the Gray conference room on the first floor. ISIS 1025 16th Ave. South. 

Recently, many applications have arisen in distributed control that require consensus protocols. Concurrently, there have been a proliferation of malicious attacks on large-scale distributed systems. This talk addresses reaching consensus in the presence of adversaries, whenever the network is itself changing due to lossy channels or mobile agents.

Recently, many applications have arisen in distributed control that require consensus protocols. Concurrently, there have been a proliferation of malicious attacks on large-scale distributed systems. Hence, there is a need for (i) the formulation of consensus problems that specify appropriate agreement and safety conditions, that are resilient in the presence of adversaries, and (ii) the design of consensus protocols for distributed control applications that solve such consensus problems resiliently despite breaches in security.

This talk presents a new consensus problem, suitable for distributed control. It is named the adversarial asymptotic agreement problem, and requires that the uncompromised agents asymptotically align their states while satisfying an invariant condition in the presence of adversaries. A low complexity consensus protocol, the Adversarial Resilient Consensus Protocol (ARC-P), which combines ideas from distributed computing and cooperative control, is introduced to solve this problem. Two types of adversaries are considered: Byzantine and malicious agents. Byzantine agents can convey different state trajectories to different neighbors in the network, whereas malicious agents must convey the same information to each neighbor. For each type of adversary, sufficient conditions are discussed that ensure ARC-P guarantees the agreement and safety conditions in both static and switching network topologies, whenever the number of adversaries is bounded by a constant. The conservativeness of the conditions is examined, and the conditions are compared to the known bounds in the literature.

Bio:
Heath LeBlanc is a fourth year PhD student in Electrical Engineering at Vanderbilt University under the direction of Prof. Xenofon Koutsoukos. His research focuses on resilient cooperative control of networked multi-agent systems, in particular, resilience to network intrusions and implementation effects. The research has applications in control of unmanned aerial vehicles, power networks, distributed robotics, and more generally, cyber physical systems. He graduated summa cum laude with the BSEE from LSU in 2007 and received the MSEE from Vanderbilt University in 2010.