Coordinated networked estimation strategies using structured systems theory

In this paper, we consider linear networked estimation strategies using the results from structured systems theory. We are interested in estimating a linear dynamical system where the observations are distributed over a network of agents. In this context, we devise both state fusion and observation fusion strategies that guarantee a stable estimator. We assume global observability, i.e., given all of the observations, the dynamical system is observable. To derive our results, we employ the genericity properties of dynamical systems that are studied in the structured systems theory. The genericity properties rely on the graphical properties of the dynamical systems and their outputs, and thus, depend on the zero and non-zero pattern of the system and output (observation) matrices. In particular, we study the generic observability of networked estimators and derive results on the topology of the agent communication graph to ensure a stable estimator. We then focus on the design of local estimator gains that results into iterative procedures to solve a Linear Matrix Inequality (LMI) with structural constraints.