A decentralized moving horizon observer for distributed implementation of centralized controllers

This paper presents a decentralized moving horizon observer (DMHO) which enables the distributed implementation of centralized controllers. The observation problem is formulated within the moving horizon observer (MHO) framework as a convex optimization problem which depends on the system dynamics and the latest noisy measurements. These measurements are distributed among the observers by an allocation algorithm which guarantees local observability. The communication topology of the observers is represented by a graph and reflected in the optimization problem by additional consensus constraints. A dual decomposition technique is used to derive a decoupled dual optimization problem which is solved by a subgradient method to yield a decentralized algorithm. Simulation results of a closed-loop four tank system demonstrate the effectiveness of the proposed method.