In order to meet the requirement of precision image stabilization system for ground-based large aperture telescopes, the control method of large aperture fast steering mirror(FSM） was studied. For the sake of solving the difficulty of system identification caused by motion decoupling of three actuator driven FSM, the transfer function model of FSM was established by combining analytical method and system identification method. According to the model, the PID controller and the model predictive controller (MPC) were designed, and the performance of the two controllers were compared by simulation and experiment. The simulation results show that the recovery speed of the MPC controller was 45 times faster than that of the PID controller after step disturbance. Under 50 Hz sinusoidal signals, due to the large inertia of FSM, the PID controller had a serious time delay, while the MPC controller could track steadily with an error of 1.224×10-6 ″. In terms of noise suppression, under the random signal with 10% amplitude noise added in real time, the noise suppression performance of the MPC controller was 13.3 times better than that of the PID controller. The experimental results show that 50 Hz sinusoidal signal could be tracked stably by the MPC controller with an error of 0.430″. The tracking accuracy of the MPC controller was 3.212 times higher than that of the PID controller. The results show that the fast steering mirror with MPC controller could satisfy the requirements of high bandwidth and high precision of the fast steering mirror.