Heterogeneous state estimation in dynamic networked systems

Distributed state estimation refers to a method for computing a local estimate of the global state in each node of a networked systems. Typically, current methods take a homogenous approach, where the same estimator is performed in each node for approximating a centralised estimation result. Yet, many practical examples indicate that the capabilities of a networked systems rise when its nodes do not only have similarities, but, to some extent, differ from each other. Inspired by this idea, this study extends current studies on homogeneous estimation and turns it into a heterogeneous solution for distributed state estimation focusing on three contributions. Firstly, the estimation problem in networked systems is reformulated so that a particular interdependency of the estimation results between any two nodes is pursued rather than aiming for similar estimation results across all nodes. Secondly, a node setup is adopted suiting heterogeneous estimation on system level, yet, still allows for nodes to employ solutions based on existing (homogenous) distributed estimators. Thirdly, the pursued interdependency of local estimation results is proven for the adopted setup, along with practical conditions that will guarantee stable estimation results in each node locally. Extended designs of this heterogeneous estimator are studied for observing the spread of a chemical compound (linear) and for detecting shockwaves on a highway (non-linear).