Generally speaking, in a satellite autonomous navigation system, it is not easy to build a state model of the practical system, which requires the information fusion algorithm having some self-adaptive capability. However, due to nonlinearity in the system measurement model, the information fusion algorithm must maintain high accuracy and robustness in a strongly nonlinear circumstance. To this end, an advanced federal adaptive unscented Kalman filter(UKF) method is proposed based on star sensor, infrared horizon sensor, agnetometer, radar altimeter and ultraviolet sensor. This method combines information from multiple navigation sensors in the federated filter, and uses an adaptive UKF algorithm to build the local filter. With this method, information coming from navigation sensors can be effectively organized and fully utilized, and the system model possesses adaptability. Numerical simulation using the proposed method is compared to that only using a conventional federated Kalman filter. The results show that the proposed method is more suitable for systems that are highly nonlinear or have inaccurat parameters, and can make navigation more accurate.