The invention discloses an interactive filtering method for satellite attitude determination, and aims to overcome the defect that constant gyro drifts cannot be estimated by utilizing the optimal-REQUEST algorithm and improve the algorithm applicability. The method comprises the following steps: step 1, acquiring sensor measurement data, including gyroscope data and star sensor data; step 2, building a state space model of a satellite attitude estimation system, as well as building an attitude K matrix; step 3, according to the state space model, the state estimate of the known moment k, and a gyro measurement value, estimating constant gyro drifts at the moment k+1 by utilizing the CKF algorithm and the optimal quaternion at the moment k, and then compensating the gyro measurement value; step 4, according to the compensated gyro measurement value, carrying out time update and measurement update by utilizing the optimal-REQUEST algorithm, so as to obtain the optimal K matrix at the moment k+1; the remaining steps of obtaining the optimal quaternion at the moment k+1. The interactive filtering method is beneficial to improvement of the estimation precision.