Distributed Innovation-Constrained State Estimation Over Sensor Networks: A State-Dependent Coefficient Parameterization Approach

The paper focuses on the distributed state estimation problem for a specific type of nonlinear systems over sensor networks. The state estimation is accomplished on each sensor node by utilizing local measurements along with network shared innovations from neighboring nodes. A state-dependent coefficient parameterization technique is employed to express the original nonlinear system in a pseudo-linear form. Additionally, a saturation-like scheme is used to limit the impact of anomalous signals during propagation and ensure the received information remains within a permissible range. The main goal of the research problem is to propose an estimation approach that restricts the estimation errors within pre-defined ellipsoids. Adequate conditions for solving this problem are provided by establishing the feasibility of a series of matrix inequalities. Further, a technique is developed to determine the locally optimal estimator parameters based on the established framework. Finally, the correctness of the results is demonstrated using a simulation example that involves the Van der Pol equation.