Abstract: Based on the analyses of the mathematical model of the inertial platform stabilization loop, an improved linear quadratic Gaussian (LQG )control method for stabilization servo circuits is put forward to overcome the disturbance from the platform. An integral element is introduced into the feedback to eliminate the stabilization error. According to the criterion of the filter convergence, the state is estimated by using Sage-Husa adaptive filtering algorithm and strong tracking Kalman filter respectively to ensure the state estimation accuracy and the tracking capability of unexpected state. The simulation and experiment indicate that improved LQG control method is accurate and capable of anti-jamming, and it relieves the accuracy requirement for the system model and noise statistical characteristics.