Firstly, this paper defines the definition and research scope of intelligent space system, and investigates the development of foreign intelligent space system under the three core characteristics of intelligent information extraction, autonomous task planning and intelligent information interconnection; Then, the development of intelligent space system in China under these three characteristics is compared and investigated, and the scientific problems existing in the intelligent space system are analyzed. Finally, the development trend of key technologies of intelligent space system in the future is proposed.

Beijing 3 satellite, successfully launched in June 2021, is a novel agile optical remote sensing satellite with the characteristics of intelligence, high agility and dynamic imaging. As one of the innovative technologies in Beijing 3, autonomous mission planning greatly makes Beijing 3 more intelligent and user friendly compared to all previous satellites. In this paper, the operating mode and payload working modes of Beijing 3 are firstly introduced. Then, the detailed design of autonomous mission planning is proposed, including the core functions and key technical requirements, the systematic design of hardware based on homemade DSP chip, the information flow design of autonomous mission planning from ground injecting original mission data blocks to onboard generating and outputting the subsystem commands, and the realization of all functional modules of autonomous mission planning software. Finally, the inorbit test results and application of the implemented autonomous mission planning in Beijing 3 during the past year are summarized. Its rapid, accurate and coordinated planning with steady performance demonstrates the rationality and effectiveness of the proposed design.

In order to deal with the dense imaging takes of observation satellite, this paper proposes an autonomous aggregation method for imaging tasks of observation satellite based on intelligent clustering. By using DBSCAN intelligent clustering method, the original tasks are clustered into multiple dense subtask sets. The aggregation type of dense subtask set is determined according to the task density, and then the dense scattered tasks are aggregated into multiple strip observation tasks or regional observation tasks. The performance of observation satellite is greatly improved. Finally, a simulation experiment is given to verify the effectiveness of the proposed method.

An intelligent framework based on deep neural networks (DNNs) is proposed to achieve the evasive impulse for spacecraft against close proximity non cooperative rendezvous. First, a double layer mathematical programming (MP) model is established to describe the evasive impulse optimization problem. Then, the input and output parameters of DNNs are carefully selected. Based on the double layer MP model, a dataset is established by using the particle swarm optimization (PSO) algorithm to obtain optimal evasive impulses under different relative states. Finally, DNNs are designed and trained, and the hyper parameters of networks are elaborately chosen by evaluating the learning performances. Simulation results indicate that well trained DNNs can calculate optimal evasive impulses with a high precision and a fast speed. Our approach can promote the intelligentization of on orbit evasion and efficiently improve the survivability of spacecraft in the orbital game.

Aiming at the complex target imaging characteristics of remote sensing images, the traditional object detection and recognition technology has the problems of low accuracy and insufficient robustness. An on orbit object detection method is proposed in this paper based on deep learning for optical remote sensing image. At the hardware level, this method uses FPGA and multi core DSP to build an on board hardware processing platform, which can update the on orbit object library and optimize the performance of the deep learning model. At the algorithmic level, this method introduces deep separation convolution based on YOLOv3 feature extraction network(Darknet 53) to effectively compress model parameters and inference computation. In the detection stage, a local re detection module is added to improve the adaptability of the algorithm to dense targets. Compared with the current object detection methods, this method has a great improvement in processing speed and accuracy, the detection accuracy of the target is higher than 90%, and the unit processing speed reaches 334.24FPS. At the same time, it supports the detection of aircraft, ships, vehicles and other typical targets, laying a foundation for model application.

To meet the requirements of high efficiency and high performance on board computing, an on orbit reconfigurable, intelligent and evolvable heterogeneous computing system architecture was proposed based on highly reliable minimum system and intelligent heterogeneous computing unit. Based on the heterogeneous computing system architecture, a reconfigurable FPGA DSP intelligent heterogeneous computing unit is designed and implemented. The heterogeneous computing unit has the function of highly reliable minimum system, high performance FPGA logic and rich DSP multi core resources, and can support up to 4 kinds of FPGA programs and 6 kinds of DSP program reconstruction. It can support the on orbit processing of multi type parameter annotation, time sharing loading to complete different tasks, and can support up to 24 kinds of program combination. Based on the FPGA DSP computing unit, a computing system for on orbit processing of remote sensing images is designed and implemented, which has preprocessing functions such as cloud detection and relative radiation correction, and on orbit processing functions such as ROI extraction based on geographic location, ROI extraction based on target detection and user processing of extracted slice images. In the future, based on this heterogeneous computing system architecture, intelligent heterogeneous computing units with higher performance will be constructed, the number and types of computing units will be flexibly configured, and more advanced space borne intelligent heterogeneous computing systems will be realized.