Stability of consensus extended Kalman filtering for distributed state estimation

The paper addresses networked estimation of the state of a nonlinear dynamical system. It is shown how, exploiting a suitable consensus approach wherein prior and novel information are dealt with in a separate way along with the extended Kalman filter linearization paradigm, the resulting distributed nonlinear filter guarantees local stability under minimal requirements of network connectivity and system collective observability. A simulation case-study concerning target tracking with a network of nonlinear (angle and range) position sensors is worked out in order to show the effectiveness of the considered consensus filter.