Aiming at the problem of fault estimation and fault-tolerant control under the constraints of actuators, the longitudinal and lateral linearized motion models of guided rocket containing the fault model were firstly established. Then, considering the time-varying uncertainty of the system parameters, the fault diagnosis and estimation methods based on neural network observers were introduced to effectively observe the degree of fault and system state parameters. By designing a robust adaptive fault-tolerant compensation controller, it solved the problem of state convergence under different degrees of actuator fault or stuck fault. Finally, three fault types of numerical simulations verified the effectiveness and robustness of the proposed algorithm for fault handling of actuators.