Abstract: For the optimal control problem of large uncertainty systems with partially unobservable states, an output feedback adaptive value iteration control method is proposed. Firstly, a state observer based on RBF (Radial Basis Function) neural network is constructed to estimate the unobservable states of the system; based on Lyapunov stability theory, an adaptive law for network weights is designed, achieving asymptotic estimation of the unobservable states. Secondly, a VI (Value Iteration) framework incorporating the RBF neural network observer is established; based on the system output and input data streams, the algebraic Riccati equation is solved to obtain the optimal control law of the system. Stability and convergence analyses are completed using the Lyapunov method, proving that the closed-loop system is uniformly ultimately bounded. Finally, for the large uncertainties and partially unobservable states generated during large-scale maneuvers of hypersonic vehicles, an adaptive robust controller is designed and simulation experiments are conducted; the results show that the designed controller can maintain system stability and excellent tracking performance even in the presence of large-scale maneuver uncertainties.