Data Aggregation in multi-agent systems in the presence of hybrid faults.

摘要

Data Aggregation (DA) is a set of functions that provide components of a distributed system access to global information for purposes of network management and user services. With the diverse new capabilities that networks can provide, applicability of DA is growing. DA is useful in dealing with multi-value domain information and often requires the agents to exchange messages with the others to reach at global information. With respect to the DA problem, the research in Partially Connected Networks (PCNs), e.g. sensor, mobile, and ad-hoc networks is very limited. This is due to the complexity of formal proofs and the fact that an agent may not have a complete view of the global network, which makes it difficult to attain the convergence properties. This complexity is compounded in the presence of message dropouts, faults, and orchestrated attacks.;Rather than providing solutions for specific forms of attacks, flaws, or defects, the research takes a unique approach to the DA problem that will be resilient against the impacts of unforeseen negative events with the potential of preventing the system to operate correctly or perform optimally. Hence, the distinction between orchestrated and accidental misbehavior would become irrelevant.;This research provides solutions to the DA problem for synchronous PCNs under fault free environment, omissive faults, and malicious behavior. Performance expressions such as network-convergence, stationary-convergence, and convergence rate have been obtained that can be beneficial in addressing the needs of the DA application at hand. The investigation has revealed that the impact of network characteristics and DA parameters are often intertwined, with the network diameter having a major role in the performance of DA algorithms.;Keywords: Consensus, Convergence, Data Aggregation, Distributed Agreement, Gossiping, Hybrid Fault Model, Partially Connected Network, Peer-to-Peer Network.