Robust Information Sharing Algorithm against Faults for Federated Filter

This paper surveys how the information sharing factors influence the robustness of the healthy local filter to the sensor failure in the federated filter (FF) with two local filters. By minimizing the failure’s influence on the healthy local filter, a robust information sharing idea is theoretically developed to improve the robustness of the local filters to the failure. An information division law based on the fault detection function of the residual Chi square test is proposed for the FF only comprising double local filters. From the viewpoint of probability, this law can make the healthy local filter achieve high resistance to the contamination from the faulty local filter. Based on the assumption that each sensor is probably more healthy than another sensor, a general method to evaluated the information sharing factors under a given fault attenuation coefficient constraint is proposed for the FF with a master filter and two local filters. Theoretical analysis shows by this method each local filter is robust to the sensor failure, which is highly advantageous for the reconstruction of the federated filter after fault isolation. Finally, simulation demonstrates this idea is effective.