LUO Yazhong , YANG Zhen , WANG Hua , ZHOU Jianping
2025, 47(4):1-9. DOI: 10.11887/j.issn.1001-2486.25040004
Abstract:Space mission design software spans the entire lifecycle of space missions and is regarded as the cornerstone of the aerospace industrial software system. Leveraging nearly three decades of technical expertise in the field of manned spaceflight, the research team from the National University of Defense Technology initiated the development of the independent industrial software ATK(aerospace tool kit) in 2020. By the end of 2024, the ATK 3.0 version was officially released, encompassing 5 major categories and 21 functional modules, including standard platform, visibility and coverage analysis, mission analysis, orbital design, and secondary development. The development of ATK has been deeply and continuously supported by China Manned Space Engineering exploring a collaborative research and development model characterized by "engineering-driven demand+centralized academic development". This software has had a significant impact in aerospace engineering, space security, and academic education, marking a solid first step toward replacing STK. The development history of ATK was reviewed, and the functional features was elaborated on. Focuses on analyzing its core technologies breakthroughs of ATK, such as large-scale software architecture design, computational kernel development, and orbital maneuver planning were selective analysed. While ATK faces challenges in enhancing its functionality and building its application ecosystem, its future development roadmap and strategic objectives were outlined.
YANG Zhen , JIN Ke , GUO Xiang , WANG Hua , LUO Yazhong
2025, 47(4):10-22. DOI: 10.11887/j.issn.1001-2486.24120053
Abstract:Orbital maneuver planning is an important part of the design of complex space missions such as rendezvous and docking, lunar exploration, etc. However, the varying configuration requirements of different missions for orbit transfer positions, maneuver magnitudes, and targeting parameters pose significant challenges to the generalized modeling and solution of maneuver planning. Regarding this problem, a generalized orbital maneuver planning modeling method based on orbital segments was proposed, which abstracts the maneuver requirements of different scenarios into orbital segments, stopping conditions, and constraints, forming a building-block spacecraft mission description model. The orbital maneuver requirements were converted into a unified nonlinear programming problem, which was subsequently solved by employing three distinct methods: differential correction, sequential quadratic programming, and intelligent optimization algorithms. A software module ATK. Astromaster was developed as a core module for ATK(aerospace tool kit) software. Simulation results show that the proposed method can achieve general orbit maneuver modeling and solving in different scenarios.
WANG Hua , HUO Chengyi , HE Junhua , YU Dateng , FANG Xuankun
2025, 47(4):23-32. DOI: 10.11887/j.issn.1001-2486.24100030
Abstract:Aiming at the lack of universal software for spacecraft RPO (rendezvous and proximity operations) mission planning, the RPO planning model and software were studied. The concept of RPO element was proposed, and an element system, which consists of four categories (rendezvous, proximity, companion flight, and departure) was constructed. RPO missions can be formed by combining multiple elements. A mission planning model for RPO elements was established. Calculation formulas for the mission control segments, stop conditions, control parameters, and constraints were provided. Based on the ATK (aerospace tool kit), a RPO mission planning tool was designed and implemented. Simulations were conducted using examples of GEO rendezvous and sun synchronous fly around, and the results show that the established RPO elements and software can effectively design RPO missions.
HUANG Hexiang , YANG Zhen , LI Jiasheng , LUO Yazhong
2025, 47(4):33-41. DOI: 10.11887/j.issn.1001-2486.24110006
Abstract:In order to solve the problem that it is difficult to identify the forced motion intention of non-cooperative targets, an intention recognition method based on BiGRU(bi-directional gated recurrent unit) network was proposed. The non-cooperative target was categorizes into five forced motion intentions: "forced round fly-around""forced drip-drop fly-around""fixed-point oscillating""line approach" and "hop approach", and the forced motion intention maneuver information dataset of the non-cooperative target was established. Based on the maneuver time series information of the non-cooperative target after entering the observation range of our spacecraft, the BiGRU network was utilized to train on the potential correlation between the time series data and the forced motion intention, so as to realize the intention recognition of the non-cooperative target. The simulation results demonstrate that the detection accuracy of the BiGRU network-based forced motion intention recognition method for non-cooperative targets achieve 98.35%. This method can improve the ability to identify the intentions of non-cooperative targets and provide a technical reference for the safety of our spacecraft in orbit.
LI Mingming , GUO Shuai , ZHU Yuehe , LIANG Yangang , DONG Min , XU Xiaosheng
2025, 47(4):42-51. DOI: 10.11887/j.issn.1001-2486.25010036
Abstract:Due to the limitations of the number of satellites, scale and manoeuvrability, it is difficult for the existing earth observation constellations to respond quickly to emergency needs of a high degree of randomness. In order to meet the demand for rapid design of emergency earth observation constellations, an emergency earth observation constellation design method was proposed. Based on the secondary development and coverage analysis function of the ATK(aerospace tool kit),a one-dimensional data hierarchical clustering method was adopted to group ground targets, and then the differential evolution algorithm was appliedto optimize the restricted Walker subconstellation configuration for the target groups, and finally the restricted hybrid Walker earth observation constellation was generated . Simulation results demonstrate that the method can rapidly generate emergency constellations compared with conventional Walker constellations and violent optimisation results, and minimize satellite deployment quantity while ensuring the effective completion of the earth observation mission.
LI Linhong , LIANG Yangang , LI Kebo , WANG Jiaxin
2025, 47(4):52-63. DOI: 10.11887/j.issn.1001-2486.24120052
Abstract:A non-coplanar flyby is recognized as an effective method for monitoring high-value target satellites in geosynchronous orbit. Taking the scenario of patrolling the geosynchronous belt with an elliptical orbit as an example, a method for solving the minimal orbital intersection distance between any two elliptic orbits was proposed, and it was proven that the optimal flyby point for the patrolling satellite is the ascending node. ATK(aerospace tool kit)was used to model the mission scenario, investigating the variations in the optimal flyby point position in J2 gravitational model. The variation law in terminal constraints (e.g., relative distance and sun phase angle at the flyby point) were studied under different local times and flyby directions. Aoptimal fuel transfer strategy satisfying the maximum transfer time and terminal constraints was proposed, and mission parameters for the mission under different gravitational model assumptions were solved by using the maneuver planning module in ATK. Mission parameters corresponding to different true anomaly angles in the two-body model were analyzed, the effectiveness of the strategy was validated, and a basis for planning multiple patrol sequences was provided.
ZHU Binyu , LI Haiyang , YANG Zhen , HE Junhua , LU Lin , ZHANG Yuhang
2025, 47(4):64-75. DOI: 10.11887/j.issn.1001-2486.25030028
Abstract:The free-return orbit serves as the preferred orbital scheme for crewed spacecraft in earth-moon transfers, yet its design involves stringent constraints and significant initial-value dependency in existing algorithms. The earth-moon transfer trajectory planning for manned lunar exploration was addressed by proposing a dual-path neural network learning method to optimize free-return orbit initialization. A dynamic model of the free-return orbit was established to analyze the characteristics of the near-earth orbital solution space. Integrating the spatial partitioning characteristics of ascending and descending orbital phase in solution spaces, a dual-path neural network architecture designed via parameter-correlated transformation was proposed to ensure the completeness of orbital solutions. Utilizing ATK.Astromaster, the earth-moon free-return orbit planning under the dual-path network learning-based initialization method was implemented and validated through simulation. The results provide an effective reference for mitigating initial-value dependency in manned lunar mission orbit design.
LI Dazi , LIU Zibo , BAO Yanyang , DONG Caibo , XU Xin
2025, 47(4):76-90. DOI: 10.11887/j.issn.1001-2486.24120028
Abstract:Successful application of reinforcement learning in decision support, combinatorial optimization, and intelligent control has driven its exploration in complex industrial scenarios. However, existing reinforcement learning methods face challenges in adapting to graph-structured data in non-Euclidean spaces. Graph neural networks have demonstrated exceptional performance in learning graph-structured data. By integrating graphs with reinforcement learning, graph-structured data was introduced into reinforcement learning tasks, enriching knowledge representation in reinforcement learning and offering a novel paradigm for addressing complex industrial process problems. The research progress of graph reinforcement learning algorithms in industrial domains was systematically reviewed, summarized graph reinforcement learning algorithms from the perspective of algorithm architecture and extracted three mainstream paradigms, explored their applications in production scheduling, industrial knowledge graph reasoning, industrial internet, power system and other fields, and analyzed current challenges alongside future development trends in this field.
WANG Zelong , WU Yuhang , LI Jian , YANG Xuan
2025, 47(4):91-110. DOI: 10.11887/j.issn.1001-2486.24120027
Abstract:Diffusion models represent a novel type of generative artificial intelligence models. Compared to traditional networks such as generative adversative networks, variational autoencoders, and flow models, diffusion models are characterized by their robust training, high fidelity and diversity in generation, and strong mathematical interpretability, and so they are widely used in fields of computer vision, signal processing, multi-modal learning and so on. Diffusion models are capable of sufficiently learning and exploring the deep generative priors from the training images, providing a novel paradigm for solving inverse problems in image processing. In order to systematically sort out the development status of diffusion model, especially the latest progress in solving the inverse problem of image processing, the research of diffusion model for the inverse problem of image processing was reviewed. The basic principle and development status of diffusion model was expounded,the main technical route of using diffusion model to solve the inverse problem of image processing and some specific application results in this direction were emphatically introduced, and the future research directions were envisioned.
REN Junkai , QU Yuke , LUO Jiawei , NI Ziqi , LU Huimin , YE Yicong
2025, 47(4):111-122. DOI: 10.11887/j.issn.1001-2486.24120032
Abstract:Long-sequence autonomous manipulation capability becomes one of the bottlenecks hindering the practical application of intelligent robots. To address the diverse long-sequence operation skill requirements faced by robots in complex scenarios, an efficient and robust asymmetric Actor-Critic reinforcement learning method was proposed. This approach aims to solve the challenges of high learning difficulty and complex reward function design in long-sequence tasks. By integrating multiple Critic networks to collaboratively train a single Actor network, and introducing GAIL (generative adversarial imitation learning) to generate intrinsic rewards for the Critic network, the learning difficulty of long-sequence tasks was reduced. On this basis, a two-stage learning method was designed, utilizing imitation learning to provide high-quality pre-trained behavior policies for reinforcement learning, which not only improves learning efficiency but also enhances the generalization performance of the policy. Simulation results for long-sequence autonomous task execution in a chemical laboratory demonstrate that the proposed method significantly improves the learning efficiency of robot long-sequence skills and the robustness of behavior policies.
LUO Junren , ZHANG Wanpeng , SU Jiongming , LI Shengqiang , CHEN Jing
2025, 47(4):123-131. DOI: 10.11887/j.issn.1001-2486.2409002
Abstract:Aiming at the dynamic force deployment problem, a multi-agent reinforcement learning strategy planning method based on SVD (Shapley value decomposition)was proposed. The reward distribution among cooperative multi-agents was explained by SVD, and the reward distribution was analysed by SVD reinforcement learning method to solve Markov convex game strategy. Secondly, based on the scenario of naval and air cross-domain cooperative confrontation, the allocation of space domain combat resources in heterogeneous multi-entity cooperative confrontation was analysed, a dynamic force deployment strategy planning model was built, and the state space, action space and reward function of the problem were designed. Finally, based on typical application scenarios, simulation experiments were organized to verify the dynamic force deployment problem with the military chess deduction system. Results show that compared with the multi-class baseline algorithm, the proposed method has excellent performance in strategic planning of dynamic force deployment, and it is theoretically interpretable. The proposed method learns the strategy of "layer upon layer interception, zone confrontation, core cover, and hierarchical breaking".
ZHU Pengming , YANG Jiaqi , LIU Peng , QIU Xuekai , DAI Wei , ZENG Zhiwen , LU Huimin , ZHOU Zongtan
2025, 47(4):132-142. DOI: 10.11887/j.issn.1001-2486.25030008
Abstract:To better investigate the complex collective behaviors in distributed collaborative robots, a multi-robot dynamic visualization research platform based on three-wheeled omnidirectional mobile robot was designed. The purpose of the platform is to provide an intuitive and flexible experimental environment for promoting the testing and development of multi-robot algorithms. The platform is composed of self-developed, low-cost, small, omnidirectional wheeled mobile robots and visual multi-touch screens that support gesture recognition and shape detection of objects, enabling the configuration of various dynamic rendering scenarios. With this platform, researchers are able to focus on the design and optimization of algorithms in multi-robot systems without being limited to specific scenarios or task settings. The robots motion performance was tested, and multi-robot algorithms were successfully tested in multiple task scenarios, initially validating the platforms effectiveness and flexibility.
LIU Haijun , ZHANG Chenxi , WANG Xiyu , CHEN Changlin , CHEN Jun , LI Zhiwei
2025, 47(4):143-150. DOI: 10.11887/j.issn.1001-2486.25010015
Abstract:To address the problem of how to faithfully map neural networks to resource-constrained embedded devices, a mixed-precision quantization method for convolutional neural networks based on layer sensitivity analysis was proposed. The sensitivity of convolutional layer parameters was measured by calculating the average trace of the Hessian matrix, providing a basis for bit-width allocation. A layer-wise ascending-descending approach was employed for bit-width allocation, ultimately achieving mixed-precision quantization of the network model. Experimental results demonstrate that compared to the fixed-precision quantization methods DoReFa and LSQ+, the proposed mixed-precision quantization method improves recognition accuracy by 10.2% and 1.7%, respectively, at an average bit-width of 3 bit. When compared to other mixed-precision quantization methods, the proposed approach achieves over 1% higher recognition accuracy. Additionally, noise-injected training effectively enhances the robustness of the mixed-precision quantization method, improving recognition accuracy by 16% under a noise standard deviation of 0.5.
2025, 47(4):151-157. DOI: 10.11887/j.issn.1001-2486.24120030
Abstract:Deep image clustering was employed to analyze the cluster structure of unlabeled image data through deep learning techniques. However, due to the absence of class labels that provide definitive information, uncertain clustering predictions may be yielded by unsupervised deep image clustering, introducing noise information that was found detrimental to performance enhancement and application development. Therefore, a clustering prediction optimization method based on alternating normalization and category-wise uniform prior was proposed to correct low confidence predictions and improve deep image clustering performance. At the same time, the method had a low degree of coupling with the model structure and training process, enabling cross-model optimization for deep image clustering frameworks. Experimental results on multiple datasets reveal that the effective clustering prediction optimization is achieved for various deep image clustering models through the approach.
YANG Wei , ZHANG Changsheng , LIU Hui
2025, 47(4):158-169. DOI: 10.11887/j.issn.1001-2486.24120025
Abstract:Given the challenges posed by photovoltaic component damage detection and the high demands placed on human and computational resources by existing detection technologies, an improved lightweight model named YOLOv8-DM was proposed on the basis of the YOLOv8n.The integration ofelectroluminescence imaging with object detection methods was implemented to achieve photovoltaic defect detection. Innovative components were introduced, including a dynamic scale feature pyramid network and an inverted residual multiscale attention mechanism, along with a Ghost module enhanced by dynamic convolution. These modifications were specifically designed to address the deficiencies observed in the YOLOv8n model regarding feature representation and multiscale object recognition, which enhanced fine-grained detection capabilities and reduced computational complexity. When evaluated on the augmented PVEL-AD dataset, the model demonstrated an improvement of 3% in recall rate and 3.3% in mAP50 compared to the baseline model, with a 34% reduction in parameter count and a 20% decrease in computational demand. The optimized architecture was validated to effectively meet the practical requirements for high-accuracy photovoltaic defect detection with lower computational costs.
ZHANG Wenxuan , SUN Fuzhen , WANG Aofei , ZHANG Zhiwei , WANG Shaoqing
2025, 47(4):170-179. DOI: 10.11887/j.issn.1001-2486.20250416
Abstract:To address the problem of inadequate self-supervised signal quality in contrastive learning models for sequential recommendation tasks, a combinatorial enumeration and time-interval contrastive learning for sequential recommendation model was proposed. The model generated enhanced sequences which preserved temporal information through time-interval perturbation-based data augmentation. A combinatorial enumeration strategy was introduced to integrate user behavior and time-interval information, constructing multi-view augmented sequence pairs. The model employed a multi-head attention mechanism to encode user behavior sequences and optimized self-supervised signals through multi-task joint training, which improved model performance. The proposed model is well-suited for scenarios with high data sparsity and uneven interaction behaviors, effectively addressing challenges in self-supervised signal modeling. Experimental results on three real-world datasets demonstrate that the model outperforms the current state-of-the-art contrastive learning models in terms of HR (hit ratio) and NDCG (normalized discounted cumulative gain).
HUANG Chengyan , ZHA Xiaoyun , DING Qunyan , HU Wei
2025, 47(4):180-188. DOI: 10.11887/j.issn.1001-2486.24120026
Abstract:To enhance the ability of large language models to generate legal documents for the power grid sector under few-shot conditions, a few-shot legal document generation method based on LLM(large language models) was proposed, integrating data augmentation and rule guidance techniques. The proposed method addressed key challenges in power grid legal document generation, such as data scarcity, high domain specificity, and the complexity of legal practice. Experimental results show that the method achieves excellent performance in generating power grid legal defense documents, significantly improving the quality and professionalism of the generated texts.
KONG Detong , LI Naipeng , LI Xinyu , LIU Chao , ZHANG Leping , HUANG Yuhao
2025, 47(4):189-196. DOI: 10.11887/j.issn.1001-2486.24100018
Abstract:Gear faults manifest in the frequency spectrum as multi-order modulation sideband clusters phenomenon center on the meshing frequency and its higher-order harmonics, spaced by the gear rotation frequency. In order to automatically focus the fault side frequency components, a method of fault side frequency cluster extraction with penalty regression was proposed. Adaptive sparse group lasso regression self-data-driven strategy was used to determine the penalty coefficient size and update the spectrum weight online to find the fault sideband clusters. Based on the sideband weight coefficients obtained from sparse group lasso regression, a new index called the sparse group lasso sidebands indicator was proposed for the health monitoring of gear transmission systems, enabling the early fault warning and location of gear transmission systems. Results analysis show that the proposed method can provide more accurate gear early fault detection and fault location.
WANG Nan , SUN Leyuan , NIU Yifeng , HUANG Wende
2025, 47(4):197-205. DOI: 10.11887/j.issn.1001-2486.24010031
Abstract:In order to solve the problem of lacking of time and space references of the autonomous navigation satellite constellation relying on inter-satellite links, the BeiDou satellite navigation system has introduced ground anchor stations which jointly constitute an integrated satellite-station network with the space satellites. With the constraints of inter-satellite link system, visibility and load quantity, the integrated assignment model of satellite-station/inter-satellite links was established. The shortest information transfer link assignment algorithm based on network layers, and the multi-objective optimization algorithm of ranging link PDOP(position dilution of precision)and network connectivity based on the simulated annealing algorithm were proposed. According to the simulation results, the reference information from the anchor station can be distributed to satellites with the fewest link hops. The PDOP of satellite ranging links is less than 1.4 and close to the lower limitation. The connectivity of the integrated network is better than 3. It is demonstrated that the integrated assignment algorithm can satisfy the requirements of fast information distribution from the anchor station to satellites, ranging for space reference maintenance and network robustness.
ZHOU Jingyuan , LEI Siyang , RAN Xianwen , TANG Wenhui
2025, 47(4):206-214. DOI: 10.11887/j.issn.1001-2486.25010013
Abstract:To investigate the mechanical properties of PTFE/Al (polytetrafluoroethylene/aluminum) reactive materials and temperature effects on their impact ignition behavior, eight PTFE/Al reactive material specimens were fabricated through controlled sintering conditions. Stress-strain curves of eight samples at varying temperatures were obtained by using a universal testing machine, while impact ignition thresholds of eight samples under corresponding temperatures were determined through drop-weight experiments. Experimental results demonstrated that sintering duration significantly influenced the elastic modulus of PTFE/Al composites, with elevated temperatures inducing material softening. At high-temperature conditions, specimens sintered for 40 min exhibited enhanced yield strength and demonstrated higher reactivity under drop-weight loading. Further analysis revealed a linear negative correlation between material strength and impact ignition thresholds under isothermal conditions, with the slope of this linear relationship exhibiting exponential decay as temperature increased. The findings of this study provide a significant theoretical foundation for the performance optimization and engineering applications of PTFE/Al reactive materials.
ZHAO Jiashan , FAN Zhenglei , ZHANG Junlong
2025, 47(4):215-226. DOI: 10.11887/j.issn.1001-2486.24120021
Abstract:Source region integration algorithm is a kind of data processing method to extract noise characteristics of each component in wind tunnel test model. To solve the interference problem of integration results caused by sound source sidelobe outside the target integration, a partial-CLEAN integration algorithm was proposed. This algorithm divided the model region into target and non-target regions, and used the CLEAN algorithm to find the strongest sound source positions at each frequency point in the non-target region. It iteratively removed interference from non-target region sound sources on the target region, thereby achieving more accurate integral results. Through simulation and wind tunnel test data analysis, the algorithm can effectively separate mutually interfering noise sources, especially sound sources of 3 kHz and below, extract more accurate target sound sources, and provide a new tool for aerodynamic noise analysis.
SONG Weixing , WU Gaiyin , MENG Lingzhou
2025, 47(4):227-238. DOI: 10.11887/j.issn.1001-2486.24100034
Abstract:Considering the practical needs of equipment utilization, equipment preparedness, and the efficiency of equipment maintenance, the relationship between equipment utilization and maintenance was analyzed. A multi-objective optimization model aimed at maximizing equipment presence rate, equipment motor-hour reserve compliance rate, equipment availability rate, and minimizing equipment maintenance costs was established, with the constraint of balancing the revenue and expenditure of motor hours. Given the high-dimensional and large-scale difference characteristics of the established models objectives, an improved NSGA-Ⅲ algorithm was designed, which proposed the crossover and mutation evolutionary operations based on DNA, to enhance the evolutionary efficiency of the population. An adaptive normalization method was improved to enhance the population diversity and algorithm convergence speed. Taking the annual equipment utilization optimization as an example, comparative experiments were conducted, which verified the feasibility and efficiency of the proposed model and algorithm. By analyzing the trend of objective function values for a large number of equipment utilization plans and the detailed indicators of plans that achieve extreme values for single objectives, different preference-based equipment utilization optimization plans were proposed.