Survivable Data Aggregation in Multiagent Network Systems with Hybrid Faults

With respect to the data aggregation (DA) survivability, the research in partially connected networks (PCN) is limited. This study investigates the DA survivability for PCNs in synchronous systems in the presence of hybrid fault modes, with the following in mind: 1) agents that run on network nodes use messages from their immediate neighbors only, i.e., no relay of information is allowed, 2) hybrid fault models are assumed; therefore, the DA algorithm is flexible to be tuned for various fault settings, 3) impacts of faults and threats rather than their sources are considered; hence, significant number of misbehaviors are reduced to a small number of fault modes, and 4) the network can tolerate any number of faults as long as the maximum number of faults encountered by each agent does not exceed a predefined threshold. The results show that the upper bound on the number of rounds to reach global convergence (agreement) and the asymptotic convergence per round depend on intertwined parameters such as precision of convergence, node degree, number of agents in the network, level of fault tolerance, and the network diameter. It is illustrated that the network-diameter has the most impact on the speed to reach global convergence.