This article proposes a resilient distributed state estimation framework for nonlinear cyber-physical systems (CPS) using sensor networks exposed to cyberattacks. The method integrates the distributed hybrid information fusion (DHIF) algorithm with a graph-based reconfiguration strategy that enables the system to detect and isolate false data injection (FDI) attacks on communication links and reconfigure the network to restore connectivity and maintain estimation performance. Each sensor node features a local estimator and an anomaly detector, allowing for fully decentralized adaptation to adversarial conditions. Upon detecting attack-induced disruptions, the system autonomously reconfigures its communication graph to maintain information flow and ensure continued system-wide collaboration. Theoretical analysis guarantees that estimation errors remain ultimately bounded despite network switching topologies. An unmanned aerial vehicle (UAV) case study under limited sensing and coordinated attacks demonstrates the effectiveness of the proposed framework in maintaining stable and accurate distributed state estimation.
Kazemi et al. (Thu,) studied this question.