This paper addresses the design of an unknown input observer for a class of generalized Hamiltonian systems subject to a locally destabilizing vector field. Conventional nonlinear state observer may lead to computationally inefficient implementations for this class of systems. To address this limitation, a reduced-order unknown input observer (RUIO) is developed to reduce the computational complexity associated with the conventional nonlinear state observer. The proposed approach employs a mixed estimation model in which the predicted and filtered states are not necessarily identical, thereby enabling a more computationally efficient observer implementation. The RUIO is then applied to the design of a secure communication system for message recovery. Additional observer designs are considered for comparison, and simulation results demonstrate the effectiveness and computational advantages of the proposed approach.

unknown input observer,Hamiltonian system,Chaotic system,secure communication