Abstract: There exist some unique problems in the operation of multi-rotor aerial manipulators: The underactuated multi-rotor platform is susceptible to external disturbances, such as gusts and turbulence; The internal system disturbances induced by the planned motion of the manipulator affect the attitude of the UAV. To solve these problems, we propose an anti-disturbance control method based on a variable centroid dynamics model. By combining the analysis method of multi-rigid-body centroid kinematics, we decompose and couple the internal and external disturbances into the UAV dynamics model, and construct a coupled nonlinear dynamics model of the quadrotor aircraft integrating the time-varying centroid offset of the manipulator, which reduces the complexity of integrated modeling while taking into account model accuracy. To address the challenge of complex parameter tuning in traditional Active Disturbance Rejection Control (ADRC), we design a parameter tuning method based on particle swarm optimization (PSO) algorithm, and apply the optimized parameters to the attitude controller of the aircraft platform. Simulation experiments show that the proposed method reduces the root-mean-square error (RMSE) between the estimated disturbance value and the actual disturbance value to 6.7%. A significant improvement in attitude angle stabilization is achieved, with the proposed method suppressing oscillation amplitude by 78% and 73% relative to the proportional-integral-derivative (PID) and nonlinear disturbance observer-based PID (NDOB-PID) controllers, respectively. Experimental results demonstrate that, in the presence of internal disturbances caused by the manipulator’s planned motion, the mean attitude angle deviation is confined to a range of 4.02° to 9.65°.