Robust median consensus cubature Kalman filter for distributed sensor networks

Yufan Zhou, Zhongliang Jing, Peng Dong, Jianzhe Huang

Abstract: Distributed estimation of multiagent systems has gained increasing attention in recent years, with consensus algorithms being one of the popular research topics. However, most existing algorithms aim to reach a consensus on the average value of all nodes, which can be significantly affected by outliers and disturbances. To achieve a more resilient distributed estimation, this paper focuses on the median value consensus strategy. Specifically, a continuous median consensus method with an adaptive parameter is presented and discretized using time-varying steps. This approach eliminates chattering errors caused by discretization and enhances convergence speed considerably. Furthermore, utilizing the proposed algorithm, a median consensus cubature Kalman filter (MC-CKF) is designed for nonlinear scenarios. Simulation results demonstrate that our proposed algorithm outperforms traditional algorithms when facing a denial-of-service attack.