Distributed H∞ Filtering for Linear and Nonlinear Systems

This paper develops a fully distributed approach for implementing H-infinity filtering over multi-agent networked systems. This is obtained by using diffusion techniques for the fusion of local filtering operations to enforce cooperation between agents and achieve a network-wide cohesive filtering operation. More importantly, we propose a diffusion-based algorithm which uses only locally observable states. Furthermore, the work includes the information formulation of the derived filtering framework. This information formulation not only provides the basis for the establishment of observability conditions but also allows for the extension of the derived filtering framework to nonlinear systems via the use of the unscented transform. Finally, the effectiveness of the derived estimation framework is demonstrated in two simulation examples.