Distributed estimation of nonlinear systems

In a classical distributed framework, we present a novel distributed observer for genuinely nonlinear continuous-time plants. A network of sensors monitors a multiple-outputs plant. Each sensor measures only a portion of the plant’s outputs, and the sensing capability is different from sensor to sensor. The assumption of strongly connected digraph on the underlying sensor network ensures robustness and direct communication paths between nodes. Moreover, incremental homogeneity assumptions on the plant embrace a very large class of nonlinear systems for which a distributed observer can be designed. The distributed observer consists of local observers associated with each sensor, asymptotically estimating the entire state of the plant only by using the local sensing capability and information exchanged through the communication network. Numerical simulations on a network of interconnected Van Der Pol oscillators confirms theoretical results. Robustness and switchi