Gabriele Oliva, Stefano Panzieri, Attilio Priolo and Giovanni Ulivi, "Characterizing Failures and Attacks in Average Consensus," International Journal of Systems, Control and Communications, vol. 6, n. 1, 2014. [url]
Consensus algorithms turned out to be effective methodologies in all those situations involving a distributed merging of information such as the collab- orative estimation performed by a set of agents. This method, however, is not able to cope with node disruptions. In this situation an agreement may not be reached or may be flawed by the lack of information. In fact, in the consensus it is assumed that the agents continue to communicate until an agreement is reached while, in many cases, this may not be verified, e.g., due to the disruption of a node. In this paper the effect of removing nodes during the consensus protocol with respect to the time instant of disconnection is addressed. Moreover, the case of spurious or malicious nodes connected to the network with the aim to interfere with the consensus is also considered. Simulations to corroborate the theoretical findings are eventually provided.