To address the problem of refined information fusion processing in multi-UAV cooperative navigation systems, a new method for information representation and fusion reasoning based on a complex network model is proposed. By leveraging graph theory, the conventional state-space navigation model is abstracted into a networked navigation model, where a (hyper) network captures the coupling constraints among UAV states. In this representation, nodes correspond to the states of UAVs, while edges or hyperedges embody information links, including prior knowledge, measurement data, and system dynamics. The unstructured navigation information is systematically transformed into a structured form via the Fisher information matrix and information vector. Based on the Cramér-Rao lower bound inequality, both centralized and decentralized optimal state estimation algorithms are derived to achieve minimum variance unbiased estimation for network nodes. The feasibility and effectiveness of the proposed model and algorithms are validated through numerical simulations in an autonomous cooperative navigation and positioning scenario involving 100 UAVs operating in an unknown environment.