When performing tasks in complex ocean areas, unmanned surface vehicles (USVs) may face partial missing of GPS signals. This paper presents a scheme for the autonomous navigation system of USV in such occasions. In case of insufficient satellites, the loose and tight integrated models of USV navigation system are established with the analysis of error model. The extended Kalman filter (EKF) algorithm is used for comparative studies on the two integrated models. Simulation results show that the loose combination degenerates into a single inertial navigation system in the insufficient-satellite environments, and the navigation error gradually accumulates with time. In comparison, the tight combination can ensure effective performance in a certain time range with relatively lower position error.