Abstract: Gyroless attitude determination system is an important research direction in the aerospace area. In this paper, a federated nonlinear predictive filtering algorithm is presented for the gyroless attitude determination system with multiple sensors. The approach combines the good qualities of both the nonlinear predictive filter (NPF) and federated filter. Moreover, this algorithm can solve the information fusion problem effectively for the system of which the attitude dynamics model error is Non-Gaussian distribution. Firstly, the equivalence relation between NPF and Kalman filter is demonstrated from algorithm structure and estimation criterion. The flow of federated nonlinear predictive filtering algorithm is analyzed and the calculation of the model error covariance is discussed. Secondly, the principles of the star sensor and Global Navigation Satellite System (GNSS) are introduced and the federated nonlinear predictive filtering model of star sensor/GNSS integrated attitude determination system is deduced. At last, simulation results using this algorithm indicate the filter accurately estimates the attitude of the spacecraft with the utilization of the star sensor and GNSS sensor.