摘要：AbstractIn this paper, we propose a distributed sensor selection architecture for a class of networked systems characterized by three distinct groups of nodes: plants, sensors and computational agents. We address the problem of fault tolerant state estimation by fusing information on the plants provided by possibly faulty sensors via a formal exploitation of trustworthiness concepts. Specifically, the scheme uses the opinions (trust) that agents have on the quality of the measurements to select the most suitable sensors for state estimation purposes. The trust is evaluated via a novel and original adaptive reputation mechanism that is in charge of intercepting those sensors that minimize anad-hoccriterion on the degradation level of the state measurements. Finally, it is formally proved that the resulting estimation scheme is always capable to select the healthy sensor in finite-time. Some numerical simulations instrumental to show the capability of the proposed reputation mechanism to efficiently select healthy sensors from a given set are finally reported.
