Krein space approach to decentralised H∞ state estimation

A decentralised Hoo state estimation algorithm is proposed for the multi sensor state estimation problems. The idea comes from the fact that the suboptimal Hoo state estimators are just Kalman filters in Krein space. By applying the alternative form ofthe Kalman filter equations, which is considered to be more appropriate for the decentralised filters due to its structure, to the Krein space state-space model associated with the suboptimal Hoo state estimation problem, the corresponding alternative form of the suboptimal Hoo state estimator is formulated. Using this alternative form, the decentralised H(XJ state estimation algorithm is developed, which is composed of the equations for the local and central fusion filters. The proposed estimator is robust against unknown but energy-bounded external disturbances and is also fault tolerant due to its inherent redundancy. In addition, the H(XJ central fusion equations, which are newly developed, can reduce the global estimation error much more effectively than the conventional central fusion equations. Numerical examples are given to verify the performance as well as the robustness and the fault tolerance of the proposed algorithm.