CN-122009533-A - On-orbit satellite load maintenance system and method based on edge perception and DRL

CN122009533ACN 122009533 ACN122009533 ACN 122009533ACN-122009533-A

Abstract

The system comprises an optical imaging module, a satellite edge AI processor, an execution mechanism driver and a thermal control device driver, wherein the optical imaging module is used for imaging the earth surface to obtain an image stream, the satellite edge AI processor is used for obtaining telemetry data of the image stream and a satellite, analyzing image quality through the on-orbit image quality evaluation model on the image stream to obtain image quality information, generating a focusing compensation instruction and a thermal control power adjustment instruction by utilizing a depth reinforcement learning method according to the image quality information and the telemetry data of the satellite, and the execution mechanism driver is used for driving the optical imaging module according to the focusing compensation instruction and driving the thermal control device on the satellite according to the thermal control power adjustment instruction. The satellite-borne edge maintenance method and device can complete maintenance in real time at the satellite-borne edge, and has the advantages of being high in response speed, good in instantaneity and good in maintenance effect, improving satellite imaging quality and prolonging satellite on-orbit service life.

Inventors

LU ZHOU
LOU HONGWEI
HAN JINBO
ZHAO WEICHAO
LI ZIRUI
TANG DAXIN
ZHANG XINLEI

Assignees

中国科学院长春光学精密机械与物理研究所

Dates

Publication Date: 20260512
Application Date: 20260324

Claims (10)

1. The on-orbit satellite load maintenance system based on edge perception and DRL is characterized by comprising an optical imaging module, a satellite edge AI processor and an actuator driver, wherein the optical imaging module is used for imaging the earth surface to obtain an image stream, the satellite edge AI processor is loaded with an on-orbit image quality evaluation model, the satellite edge AI processor is used for acquiring telemetry data of the image stream and a satellite, analyzing the image quality through the on-orbit image quality evaluation model on the image stream to obtain image quality information, generating a focusing compensation instruction and a thermal control power adjustment instruction according to the image quality information and the telemetry data of the satellite by using a depth reinforcement learning method, and the actuator driver is used for driving the optical imaging module according to the focusing compensation instruction and driving a thermal control device on the satellite according to the thermal control power adjustment instruction.
2. The edge-aware and DRL-based on-orbit satellite payload maintenance system according to claim 1, wherein the image quality information includes sharpness and/or imaging signal to noise ratio, the image quality information further includes noise distribution and/or spatial distribution of outlier pixels in the thermographic image, and the telemetry data includes temperature data, radiation data and current data.
3. The edge-aware and DRL-based on-orbit satellite load maintenance system according to claim 1, wherein the satellite edge AI processor is configured to obtain weight update information for the ground station and update the on-orbit image quality assessment model thereon accordingly.
4. The edge-aware and DRL-based on-orbit satellite load maintenance system according to claim 1, wherein the operations of generating focus compensation commands and thermal power adjustment commands using a deep reinforcement learning method are performed while the satellite is in the observation region.
5. The edge-aware and DRL-based on-orbit satellite load maintenance system according to claim 1, wherein the states of the deep reinforcement learning method include temperature, imaging contrast, noise variance, cumulative radiation, and the actions of the deep reinforcement learning method are focus compensation commands and thermal power adjustment commands.
6. The edge-aware and DRL-based on-orbit satellite load maintenance system according to claim 5, wherein the focus compensation command is a focus motor compensation pulse number and the thermally controlled power adjustment command is a current adjustment value of a thermoelectric chiller.
7. The edge-aware and DRL-based on-orbit satellite load maintenance system according to claim 1, wherein the deep learning algorithm employs a near-end policy optimization algorithm.
8. The on-orbit satellite load maintenance system according to claim 1, wherein the generating focus compensation commands and thermal power adjustment commands according to the image quality information and the telemetry data of the satellite by using a deep reinforcement learning method comprises the step of fusing the acquired image quality information and the telemetry data of the satellite to form a reinforcement-learned state space.
9. The on-orbit satellite load maintenance system based on edge perception and DRL according to claim 1, further comprising a conversion module for converting dark current drift of a photosensitive chip caused by temperature change into a variation of bias current, wherein the input of the satellite edge AI processor comprises the variation of bias current caused by temperature increase of the space environment where the satellite is located, and the satellite edge AI processor is specifically configured to analyze image quality based on the image flow according to the variation of bias current and by the on-orbit image quality evaluation model thereon to obtain image quality information.
10. The on-orbit satellite load maintenance method based on edge perception and DRL is characterized in that a satellite is provided with a satellite edge AI processor, the satellite edge AI processor is loaded with an on-orbit image quality evaluation model, and the method comprises the following steps: the optical imaging module is used for imaging the surface of the earth to obtain an image stream; Acquiring the image stream and telemetry data of a satellite by a satellite edge AI processor, and analyzing image quality through the on-orbit image quality evaluation model based on the image stream to obtain image quality information; the satellite edge AI processor generates a focusing compensation instruction and a thermal control power adjustment instruction by using a depth reinforcement learning method according to the image quality information and telemetry data of the satellite; The actuator driver drives the optical imaging module according to the focusing compensation instruction; the actuator driver drives a thermal control device on the satellite according to the thermal control power adjustment instruction.

Description

On-orbit satellite load maintenance system and method based on edge perception and DRL Technical Field The disclosure relates to the technical field of remote sensing satellites, in particular to an in-orbit satellite load maintenance system and method based on edge perception and DRL. Background With the increase of the on-orbit running time of the high-resolution remote sensing satellite, the optical load imaging system faces serious space environment challenges, including high-energy particle radiation to generate hot pixels of an image sensor (CMOS/CCD), severe temperature difference to cause defocusing caused by thermal expansion and cold contraction of an optical lens group, and mechanical abrasion of a focusing mechanism. Current satellite load maintenance relies primarily on "ground station monitoring mode". Satellites periodically transmit telemetry data of the load (e.g., temperature, current, historical imaging quality parameters) to the ground station via the downlink. The ground engineer uses the offline model to analyze, judge the load health status, and downlink control instructions (such as adjusting focus step and modifying thermal threshold) in the next transit window. In existing schemes, the processing of the sensor signal generally follows signal amplification-analog to digital conversion (ADC) -data packet transmission. This approach is similar to the sensor parameter acquisition procedure, relying on complex analog circuitry and ADC conversion. The prior art has the following disadvantages: (1) The real-time performance is poor, namely, the satellite-to-ground communication has serious window period limitation and transmission delay, and the sudden load performance degradation (such as instantaneous imaging abnormality caused by radiation) cannot be responded immediately; (2) The original high-definition image has large data volume, and cannot be completely returned to the ground for quality evaluation, so that the sample of maintenance decision is insufficient, and the maintenance effect is affected. Disclosure of Invention Based on the above, it is necessary to provide an in-orbit satellite load maintenance system and method based on edge perception and DRL, aiming at the problems of poor real-time maintenance of satellite load and to improve maintenance effect. In order to solve the problems, the present disclosure adopts the following technical scheme: The on-orbit satellite load maintenance system based on edge perception and DRL comprises an optical imaging module, a satellite edge AI processor and an actuator driver, wherein the optical imaging module is used for imaging the surface of the earth to obtain an image stream, the satellite edge AI processor is loaded with an on-orbit image quality evaluation model, the satellite edge AI processor is used for acquiring telemetry data of the image stream and a satellite, analyzing the image quality through the on-orbit image quality evaluation model on the image stream to obtain image quality information based on the image stream, generating a focusing compensation instruction and a thermal control power adjustment instruction according to the image quality information and the telemetry data of the satellite by using a depth reinforcement learning method, and the actuator driver is used for driving the optical imaging module according to the focusing compensation instruction and driving a thermal control device on the satellite according to the thermal control power adjustment instruction. In a preferred embodiment, the image quality information includes sharpness and/or imaging signal-to-noise ratio, the image quality information further includes noise distribution and/or spatial distribution of outlier pixels in the thermogram, and the telemetry data includes temperature data, radiation data, and current data. In a preferred embodiment, the satellite edge AI processor is configured to obtain weight update information for the ground station and update the on-orbit image quality assessment model thereon accordingly. In a preferred embodiment, the operation of generating focus compensation commands and thermal power adjustment commands using a deep reinforcement learning method is performed while the satellite is in the observation region. In a preferred embodiment, the states of the deep reinforcement learning method include temperature, imaging contrast, noise variance, accumulated radiation, and the actions of the deep reinforcement learning method are focus compensation instructions and thermal control power adjustment instructions. In a preferred embodiment, the focus compensation command is a focus motor compensation pulse number, and the thermal control power adjustment command is a current adjustment value of a thermoelectric refrigerator. In a preferred embodiment, the deep learning algorithm employs a near-end policy optimization algorithm. In a preferred embodiment, the generating the focus compensation command and the ther