A Self-learning Control Strategy for Vibration Isolated Control Moment Gyroscope

In order to improve the imaging resolution of remote sensing satellites, one effective method is to isolate the vibration generated by the onboard Control Moment Gyroscope (CMG). However, vibration isolators have severe impact on the system model of CMGs, which significantly degrade its performance on attitude control. In this paper, a nonlinear virtual control object model based on neural network and a selflearning compensation controller are integrated in the original control scheme to achieve strong robust control for the system. The control performance of this strategy is verified by experiments on a real CMG system.