Policy optimization for vibration isolator stiffness control during agile attitude maneuvers

Abstract

With the increasing demand for high-resolution remote sensing images, both the image quality and the image acquisition efficiency should be concerned for an agile remote sensing satellite. Effective vibration isolation is essential to ensure image quality since the attitude vibration could cause line-of-sight jitter and image blur. To improve the image acquisition efficiency, the capability of performing agile attitude maneuvers is another indispensable requirement of the satellite. These requirements mean an extreme level of attitude stability and vibration isolation from onboard disturbances. However, they have conflicting demands on the connection stiffness between the camera and the satellite bus, i.e., high on the former but low on the latter. By varying its stiffness, this paper develops a dynamic stiffness control scheme for the camera’s vibration isolator, whose variable stiffness policies are learned from a reinforcement learning methodology with a feature of low online complexity. The proposed policy optimization algorithm for stiffness control has such merits as simultaneously optimizing both vibration isolation and attitude stabilization. Furthermore, it possesses an adaptive ability to deal with pervasive onboard disturbances. Simulation experiments demonstrate that the learned variable stiffness policies could strike a favorable balance between the agility and the image quality.