Learning Based K-Band Microwave Ranging System Time-Delay Control for Earth Gravity Measurements

For the purpose of precise time-delay control for the Space-based gravity field measurement (SGFM) K band microwave ranging (MWR) system, an online reinforcement learning (RL) based thermal system design is used in this paper, with thermal control actuator realized by using passive isolation heating cubicle for antenna and patched heating device around signal process unit payload. Aiming to the better time-delay stability performance with minimal fluctuation of payload temperature under condition of thermal system disturbance in space, active thermal control of model-free RL method with actor/critic network is used for the balance of ranging payload thermal states and feedback control. Experiment results demonstrated the effective of proposed method with 25% improved ranging accuracy compared with traditional nominal control.