In order to push forward “The Third Offset Strategy” and enhance the ability of globally integrated operations, the U.S military proposes all-domain command and control (JADC2). JADC2 owns 14 abilities in 4 aspects such as joint all-domain situation cognition, joint all-domain AI-aided decision, joint all-domain command and action control as well as joint all-domain support. In order to systematize these core abilities in JADC2, the U.S military develops supporting techniques such as advanced system structure, operation cloud, military internet of things, military 5G communication, big data and supporting systems from the U.S military multiservice. In the future, the U.S military will promote the ability construction of JADC2 using the ways of four-side developing, combat verifying and AI energizing.

The Project Convergence, which the U.S. Army has been promoting in recent years, is a core experimental project for its response to great power competition and the strategic transition from Multi-Domain Operations(MDO) to Joint All-Domain Operation(JADO). It aims to dismantle the boundaries of traditional operational domains and build a capability for timely strikes across all domains through technological innovation, capability recombination, and cross-domain coordination. Firstly, the concept and positioning of the Project Convergence are clarified, and the process and characteristics of its cross-domain integration experiment and exercises are elaborated. Furthermore, it analyzes the main development trends of the project, and concludes with insights relevant to our military’s strategic planning.

In view of the decomposition and planning of missions at sea, this paper proposes a reverse mission decomposition and parameter modeling method based on the killing chain. Based on the killing chain combat theory, the four core task modules of search, tracking, strike and communication have been constructed, the functional boundaries and internal logic of each task type have been clarified, a unified task unit model has been built, and the operational sub-tasks have been formally modeled and algorithm flow designed. Adopt the reverse task decomposition mechanism, aim at the requirements of strike effect, and push back the parameters and resource conditions required for the front task step by step to ensure the consistency of the task chain parameters and the logical closed loop of the combat process. By constructing task decomposition assumptions and task instance decomposition, the advantages of this method in terms of task coverage, resource scheduling efficiency and execution accuracy are verified, and effective theoretical and technical support is provided for the coordinated operation.

The detection and recognition of small targets in complex naval combat scenarios is a critical component of achieving intelligent perception in modern naval operations. This paper provides a systematic review of the development of small-target detection technologies in such environments. This paper examines key technological advances in detection and recognition, including performance under adverse weather conditions, ship-type identification, and multi-sensor and multimodal fusion approaches. Based on battlefield requirements, this paper also explores potential future directions. Under the in-depth empowerment of artificial intelligence technology, the small target detection and recognition technology in complex naval battle scenarios is accelerating its evolution towards automation and expansion towards cluster collaboration. Meanwhile, its level of intelligence has been significantly improved, laying a solid technical foundation for the realization of all-weather and all-dimensional maritime information perception capabilities.

With the rapid development of information perception methods, the volume of multi-source heterogeneous data has increased explosively. Traditional methods exhibit obvious limitations in handing inconsistent spatio-temporal scales and information redundancy. In this paper, the principles of spatiotemporal grid coding and its application are introduced and an innovative unfied spatiotemporal grid based data association analysis method is proposed to conduce research on spatio-temporal information extraction, unified encoding and identification of spatio-temporal grid data, cross-scale spatio-temporal data sharing and dynamic correlation analysis. Our method constructs a spatio-temporal grid framework for unified alignment of heterogeneous multi-source data spatial coordinates, temporal baseline synchronization, and attribute correlation with behaviors. In addition, a spatio-temporal grid framework is constructed to align spatial coordinates of multi-source heterogeneous data, synchronize temporal baseline, and correlate attribute behaviors. The result of simulation of typical low-altitude airspace planning for Unmanned Aerial Vehicles verifies that our proposed method is enable to effectively enhance the management and utilization efficiency in classification, integration, and correlation analysis of multi-source heterogeneous data.

Based on the three threat characteristics of future maritime air swarms—"low detectability, tactical complexity, and cost-effectiveness of scale"—this paper analyzes three core requirements for constructing an anti-swarm system: " all-domain multidimensional perception, rapid dynamic decision-making, and cost-effective interception." Accordingly, it constructs an anti-swarm system featuring "multidimensional intelligent detection, dynamic agile engagement, and layered efficient interception." The paper establishes a detection system centered on "wide-area early warning, precise identification, and continuous tracking" to overcome traditional detection blind zones. It builds an engagement system focused on "intelligent situation assessment, dynamic networking, and resilient survivability" to accelerate the kill chain. Furthermore, it develops an interception system centered on "pre-emptive interception, low-cost kill, and efficient allocation" to achieve cost-effective interception. This provides a systematic framework and technical pathway for developing future naval warfare systems to counter air swarms.

In light of the fact that modern warfare places higher demands on the scientificity and timeliness of command and control, it is proposed that intelligent operational mission planning is the key breakthrough to solve the problems in the field of command and control. Firstly, the closed-loop process system of intelligent operational mission planning is summarized, and the key issues and essential contents to be addressed in each link are elaborated. Then, based on the closed-loop process system, a four-layer five-dimensional closed-loop model architecture for intelligent operational mission planning is put forward. Finally, the supporting methods, key technologies and engineering implementation key points of this model architecture are expounded.

To address challenges in intelligent mission planning, including subtask decomposition, battlefield event ambiguity, and ethical trade-offs, a system architecture based on human-machine cognitive collaboration is proposed. By deconstructing cognitive processes into five dimensions (situational awareness, assessment, plan generation, evaluation, and action control), a capability complementarity model is established, optimized through dynamic weight allocation. The three-layer architecture comprises: a data support layer integrating operational rules, historical cases, and psychological monitoring for multi-source fusion; an intelligent processing layer utilizing knowledge graphs and hybrid reasoning engines for situational deduction, ethical assessment, and innovative planning; and a human-machine collaboration layer enabling bidirectional cognition via neural enhancement, ethical constraints, and dynamic task negotiation. Some reference for improving the quality and efficiency of human-machine collaboration mission planning system can be concluded from the study.

The high concurrency and tight coupling of elements in cross-domain collaborative operations complicate mission decomposition. Two key challenges are the lack of smoothness in the decomposition process and the handling of uncertain missions. To address these, this paper proposes a multi-tiered mission decomposition method based on an extended HTN (Hierarchical Task Network) framework. HTN, which recursively breaks down complex missions into subtasks, offers structured modeling and adaptability to dynamic battlefields. This study extends HTN by integrating a knowledge base, using formation hierarchy as a key element. A knowledge base matching and evaluation method is also introduced to address uncertain missions caused by unclear intelligence and complex conditions. Validation with specific operational scenarios shows that the proposed method improves decomposition smoothness, handles uncertainty effectively, and enhances mission adaptability and executability.

In future equipment testing, real-time data collection, processing, and storage present characteristics such as large data volume, redundancy, and poor intuitiveness, which pose challenges to test command and guidance. The article summarizes the domain knowledge system for experimental command and guidance requirements, and proposes a domain knowledge graph construction method based on command and guidance data. Using the experience and knowledge of experts, a knowledge system for equipment testing command and adjustment was constructed as the initial ontology of the knowledge graph in this field. The key technologies involved in knowledge graph construction, including knowledge extraction and fusion, were explained in detail, and the specific application of this domain knowledge graph in assisting command and adjustment decision-making was demonstrated. The domain knowledge graph oriented towards command and guidance can provide support for situation analysis and system situation generation, thereby more effectively serving equipment testing command and guidance work.

With the advancement of multimodal large model (MLM) technology in capabilities such as information comprehension, logical reasoning, and content generation, its application in the military domain has reached a stage of maturity. This paper proposes a deception countermeasure architecture for command information systems based on MLM technology. First, the necessity of architectural research is clarified by examining the relationship among command information systems, military deception, and MLM technology. Subsequently, the deception countermeasure architecture and specific strategies are detailed at both the holistic and subsystem levels. Analysis demonstrates that the proposed architecture offers significant advantages in real-time performance, scalability, and flexibility. It can effectively disrupt enemy decision-making processes and prevent enemy deception operations, thereby providing engineering insights and theoretical support for the intelligent transformation of command information systems.

In response to the demands of large-scale cross-regional networked tactical confrontation flight simulation training, an intelligent simulation platform architecture based on cloud-edge-end collaboration has been designed. The architecture supported by big data and centered around intelligent large models, intelligent agents and real-time simulation models, enables multi-modal operation integrating peacetime and training running modes through cross-regional cloud-edge-end resource integration, dynamic reuse, collaborative simulation and virtual-real fusion. Based on an analysis of the functional requirements of this intelligent simulation platform, the study focuses on designing its hierarchical architecture, network structure, and synchronized simulation strategies. It investigates critical aspects including cloud-edge-end simulation task allocation, operational modes, and application scenarios. Three key technical challenges are explored including integration of cloud-based XR and AI technologies, cloud-based convergence of big data, large models and real-time simulation, and real-time interactive cloud-edge-end collaborative simulation. This work provides a reference model for constructing a new-generation intelligent simulation platform that supports all-scenario, full-system, multi-element training, thereby advancing intelligent transformation in flight simulation training domains.

The innovative development of large language models in the field of natural language processing provides new ideas for the application of artificial intelligence in the military. At present, the information service work of our army is carried out according to the traditional service concept, the submitted documents are reviewed and screened before being published. However, the ability of officers and soldiers to transform military specialized content into official documents such as news and special issues generally needs to be strengthened. This article aims to address the current situation by collecting military news from open-source channels and constructing a dataset containing over 5 000 pieces of important information and news. Using the Chatglm2-6b model from Tsinghua University as the base model, the retrieval-augmented generation method is used to improve the military domain information cognition ability of the large language model, the prompt engineering is used to optimize the document generation containing professional military knowledge. Through model evaluation and case analysis, the feasibility of the experiment has been verified, which can further support the generation of military text content and improve the efficiency of information service work in the future.

Aiming at the field of military human resources data in the era of big data, this paper proposes a military human resource big data technology combining knowledge graph and large language model (Mhr-KL). This technology eliminates data silos by constructing multi-source knowledge graph, and introduces large language model to improve intelligent interaction ability, so as to solve the problems of difficult data application and low degree intelligent interaction in the field of military human resources. The feasibility and accuracy of this technology are proved by the data island inventory test, intelligent question answering performance verification and comparative analysis in the test environment.

During the application of unmanned surface vehicles (USVs) in natural sea areas, issues such as low-quality optical image imaging, localized target features and depth-directional distortion caused by side-viewing angles, lens deformation due to saltwater droplets, and the coexistence of extremely large and small targets occur. These phenomena lead to a decline in the performance of deep learning-based image target detection algorithms, resulting in high rates of missed detections and false alarms. Consequently, this causes the collision avoidance decision-making algorithms to diverge or trigger frequent emergency alarms. To enhance object detection accuracy for USV optical images, this paper proposes an optimized framework based on the YOLOv8 model. First, the Copy-Paste algorithm and unpaired image style transfer algorithm are introduced to generate multi-view target images and low-quality images caused by adverse environmental conditions, addressing the imbalance in the proportion of low-quality images in the dataset. Second, a dedicated detection head for small targets is added to the detection network, and the loss function is optimized to enhance the detection capability for small targets while preserving the original model’s performance on larger objects. The newly developed object detection model, trained on the augmented dataset, achieves an average precision of 96.2% on the test set.

Systematized equipment, as a crucial prerequisite for informatized warfare, represents a major feature of modern weaponry. To address the issues in systematized equipment testing, such as the lack of standardized norms, insufficient environmental simulation capabilities, and deficiencies in systematic evaluation, this paper first elaborates on the concept of systematized equipment. It then analyzes the current state of systematized equipment testing and evaluation both domestically and internationally, focusing on testing models, methodologies, and capabilities. A comparative study is conducted to identify the existing problems and gaps in China’s systematized equipment testing and evaluation. Finally, in light of domestic systematized equipment testing requirements, particularly those related to combat readiness and informatization assessment, this paper proposes solutions and key measures to tackle the challenges in China’s systematized equipment testing and evaluation. The findings hold significant reference value and practical importance for advancing research in systematized equipment testing and evaluation technologies.

As one of the core elements of the naval battlefield, situational awareness plays a vital role in supporting commanders to grasp battlefield information in an all-round way and ensuring the efficient execution of command decisions. Starting from the situation of future wars, this article studies and analyzes the key intelligent projects in the field of situational awareness of the US Navy under the guidance of the new combat concepts of the US Navy, the development strategy of artificial intelligence of the US military, and combat concepts. It also summarizes its key technologies and suggestions for the application of artificial intelligence technology in the field of situational awareness by the navy.

In response to the issue of assessing the decision-making superiority of a System of Systems (SoS), this study focuses on capability packages and kill chains, proposing a system decision-making optionality concept framework and a decision-making optionality measuring model. Corresponding calculation models for the availability of capability packages and the effectiveness of kill chains within the model are also developed. The research findings of this paper are innovative in the measurement of SoS superiority, especially decision-making superiority, and can be used to assess the combat potential of mosaic SoS and combat target analysis.

Core technologies for online mission planning of hypersonic glide vehicles—rapid maneuverability assessment and autonomous trajectory planning—are ushering in new development opportunities with breakthroughs in artificial intelligence and onboard computing capabilities. By systematically reviewing research progress and trends of four representative methodologies in autonomous trajectory planning and footprint determination (including drag acceleration profile methods, quasi-equilibrium glide condition solutions, rapid optimization techniques, and deep reinforcement learning approaches), this study conducts in-depth comparative analysis. Finally, prospects are provided for key technical directions and challenges in rapid maneuverability evaluation and autonomous trajectory planning, focusing on high-precision intelligent dynamic modeling, collaborative autonomous trajectory generation, and physics-informed neural networks, aligned with the evolutionary trends of glide vehicle technologies and AI advancements.

To address the limitations of traditional cybersecurity technologies in responding to complex cyberattacks and the challenges faced by large language model (LLM) in cybersecurity applications (e.g., hallucinations, outdated knowledge bases, and insufficient interpretability), this paper proposes a construction method for a cybersecurity large model based on the Retrieval-Augmented Generation (RAG) architecture. Through three stages—functional analysis, architectural design, and technical implementation—this approach completes the "4-layer, 1-module" construction and practical deployment of the RAG system, bridging the gaps in applying generic LLM to cybersecurity domains. Furthermore, this study innovatively introduces mechanisms including parameter adaptive optimization, metadata-filtered retrieval,dynamic permission management and multi-round evaluation feedback to ensure the model’s usability, trustworthiness and security.

UAV swarms are widely used in tasks such as personnel search and rescue, as well as military reconnaissance. In order to improve the efficiency of unmanned cluster in carrying out large-scale reconnaissance tasks, a multi-objective optimization model is constructed to minimize the flight time and maximize the detection revenue for the task allocation problem of UAV cluster with different sensors. By constructing integer task encoding and a population initialization method based on Voronoi partitioning, the quality of the initial solution is improved, and the genetic method in NSGA-II algorithm is restricted to shorten the optimization time. This algorithm can provide a set of non-dominated solutions, allowing for the selection of the shortest flight time or maximum profit plan based on preference. To cope with large-scale damage, an initial population is generated based on local task flow rules to achieve rapid task optimization. Simulation results show that compared to the original algorithm, the improved algorithm has significant advantages in task allocation and damage reconstruction of large-scale unmanned clusters.

The use of UAV air raid has become a trend, especially through flexible tactical means and cluster attack, which has gradually become the basic style of air raid in various countries. Aiming at the threat of UAV cluster air raid to field air defense, the characteristics of UAV cluster air raid are analyzed, and the difficulties of field air defense against UAV cluster air raid are discussed. Aiming at the realistic threat of UAV cluster air raid, this paper puts forward the countermeasures of field air defense against UAV cluster air raid from four aspects: multiple comprehensive early warning detection, fast and efficient command and decision-making, multiple methods of comprehensive interception and protection, and research and development of new anti UAV equipment, in order to provide reference for future field air defense against UAV cluster air raid.

To investigate the effectiveness of active jamming against terminal guidance radar, this paper conducts theoretical modeling and analysis of the jamming capabilities. Firstly, based on the Linear Frequency Modulation (LFM) signal and Pulse-Doppler processing mechanism, models for the radar’s transmitted signal and the received target echo signal are established. Secondly, mathematical models for both barrage jamming and deception jamming are constructed, analyzing their respective mechanisms and limitations. Finally, jamming optimization is analyzed from three dimensions: polarization, time-frequency characteristics, and frequency agility. The study demonstrates that: (1) Dynamically maintaining the polarization mismatch angle within the range of 60°to 120°yields optimal jamming effectiveness; (2) Employing a "comb spectrum+random time delay" structure in the jamming signal makes it difficult for the radar to extract continuous features; (3) A jamming to radar agile bandwidth ratio greater than 1.2 is the critical threshold for jamming effectiveness. The results validate that regulating the polarization mismatch angle, matching time-frequency characteristics, and optimizing the agile bandwidth ratio are key pathways to enhancing jamming effectiveness.

Cooperative game theory provides a systematic mathematical modeling tool for cross domain collaborative combat by studying the formation of alliances, distribution of benefits, and collaborative optimization mechanisms. This paper systematically explains cooperative game theory and proposes a military alliance game system framework, analyzes its application mechanism in scenarios such as joint firepower strikes, integrated air and missile defense, and network electromagnetic countermeasures, and reveals its optimization effect on multi domain collaborative efficiency. Further explore the technological challenges of incomplete information and security constraints in dynamic battlefield environments, and propose future research directions for the integration of dynamic game theory, robust optimization, and artificial intelligence. This theory is a key capability support for building an intelligent combat system and enhancing the combat effectiveness of cross domain systems, and has important military application value.

Under the accelerated evolution of the Joint Electromagnetic Spectrum Operations (JEMSO) concept, the U.S. Navy has intensified its focus on theoretical research and practical implementation of distributed electronic warfare to achieve ship-aircraft collaborative electromagnetic denial and counter-electronic reconnaissance. This includes heightened emphasis on coordinated countermeasures against electronic reconnaissance threats through collaboration between surface vessels and manned/unmanned aircraft. Based on fundamental methodologies of ship-aircraft collaborative electronic warfare, this study systematically analyzes the primary equipment and operational tactics employed by the U.S. Navy. It further explores potential application scenarios and evaluates the anticipated effectiveness of counter-electronic reconnaissance. Simulation results validate that ship-aircraft collaborative electromagnetic denial significantly reduces both the interception range and interception probability of electronic reconnaissance aircraft. These findings provide critical insights for theoretical research, technological development, and combat applications of collaborative electronic warfare in China’s naval forces.

Urban warfare has become an important form of warfare that cannot be avoided in modern war, and the battle space has changed to multi-domain warfare, cross-domain warfare and all-domain warfare. In order to fully obtain multi-domain and multi-dimensional information of urban environment, effectively support urban combat decision-making and multi-arms coordinated command, a multi-domain information fusion technology system of UAVs for urban combat is proposed. The key technologies of UAV cluster collaborative deployment and networking, autonomous collection of complex urban environmental data, information fusion and spectrum completion across physical space and electromagnetic space, and multi-domain joint spectrum situation evolution prediction are analyzed. It can provide integrated "physical space+electromagnetic space" intelligent visual services for urban combat, urban spectrum intelligent management, and urban wide-area electromagnetic situation assessment and decision-making, effectively enhance the intelligent level of cross-domain information fusion representation in complex urban environments, and provide technical support for urban military and civilian electromagnetic situation intelligent monitoring, electromagnetic space cognition and control.

To address the challenge of designing and applying command and control architectures for operationally responsive space launches, the typical three-tier and four-tier architectures are studied. Based on entropy theory, a system entropy model for evaluating these architectures is constructed by integrating weighted timeliness entropy and quality entropy. The findings demonstrate that the three-tier architecture is superior when the weight coefficient of timeliness entropy (α) is less than or equal to 0.6, whereas the four-tier architecture proves more effective when α is greater than or equal to 0.67. Within the critical interval of 0.6 <α< 0.67, a threshold value u for the number of operational units (m) is identified: the three-tier architecture is preferable if m < u, and the four-tier architecture becomes optimal if m ≥ u. Furthermore, the threshold u decreases as α increases. Corresponding unit employment strategies, namely coupled utilization and streamlined utilization of combat units, are proposed for different launch modes and initial states of the launch vehicle. The models and conclusions can provide a theoretical foundation for decision-making in operationally responsive space launch command and control and its practical application.

To address the issues of insufficient global optimality and robustness in traditional mission planning for unmanned aerial vehicles (UAVs) in urban disaster relief scenarios, this paper proposes a multi-heterogeneous UAV static task allocation algorithm based on improved genetic operators. By enhancing genetic operators, designing a functionally coupled chromosome encoding strategy, introducing adaptive crossover and mutation mechanisms, and embedding a path planning module into the task allocation algorithm, the accuracy of the task allocation cost function calculation is improved. This effectively overcomes the drawbacks of traditional methods, such as inaccurate allocation results due to path cost estimation deviations, as well as poor global optimization and robustness.

With the increase in data volume, dimensionality, and the number of optimization objectives, the conflicts between objectives intensify, making the solution of multi-objective optimization problems increasingly complex. This is especially true for high-dimensional heterogeneous multi-objective optimization problems, where the difficulty of solving them increases significantly. In this paper, we propose a maximum entropy-based reference vector-guided evolutionary algorithm aimed at solving many-objective optimization problems. By combining reference point strategies with evolutionary algorithm search mechanisms, the proposed method achieves complementary cooperation between the ideal and worst reference points, thereby improving the efficiency of optimization. The algorithm relies on a set of adaptively selected reference vectors and optimizes them using Bayesian maximum entropy, focusing on balancing diversity and convergence during the optimization process. Through comparative experiments on several benchmark problems, the proposed K-RVEA algorithm demonstrates significant advantages, verifying the feasibility and effectiveness of the method.

Shipborne small-caliber weapons play an irreplaceable role in maritime rights enforcement and patrol missions conducted by Coast Guard vessels. Given the wide variety of existing small-caliber weapons, some models must be phased out based on the functional requirements and developmental goals of Coast Guard operations, while others require further improvements in tactical and technical performance. Consequently, it is imperative to establish a scientific evaluation method for assessing the operational effectiveness of these weapons. This study proposes an equipment effectiveness evaluation model based on the AHP-entropy weight method and fuzzy comprehensive evaluation. The model constructs an index system covering reliability, maintainability, combat capability, environmental adaptability, human-machine interaction, and cost-effectiveness. By fully utilizing objective factors from secondary indicators and combining subjective and objective weights, the evaluation results better reflect practical conditions. Focusing on three types of weapons, the paper compares their comprehensive operational effectiveness through case analysis. The findings offer theoretical references for decision-making in the configuration and modernization of shipborne small-caliber weapons.

In response to the problems of delayed response and inefficient resource scheduling in traditional military airport runway emergency repairs, this study proposes a digital twin technology-based emergency repair support situation analysis strategy for airport runways. By constructing a digital twin model of airport pavement, real-time perception of the destructive situation can be achieved. By utilizing the operations of command personnel to intervene in the light-weighting process of the 3D model of the digital twin airport pavement, the light-weighting efficiency is improved based on the improved 3D model light-weighting algorithm QEM, and the generation process of emergency repair and construction plans for the airport pavement is optimized. This breaks through the limitations of traditional static contingency plans and improves the operational efficiency of the emergency repair system. At the same time, it can provide effective data support for emergency repair support command personnel to plan repair plans, improve overall resource utilization, and provide reliable technical support for the rapid recovery of airport support capabilities during wartime.

In the process of bullet design, the parameters of the interior trajectory, exterior trajectory, and terminal trajectory are interdependent, necessitating a comprehensive consideration of the entire trajectory process. To address both overall performance and design efficiency, a multi-objective optimization method based on simulated annealing and particle swarm optimization is proposed. A comprehensive ballistic calculation model is established, incorporating warhead characteristics, internal trajectory, aerodynamic parameters, external trajectory, and terminal trajectory. Twelve structural parameters are selected as optimization variables.The optimization objectives are set as travel distance, landing kinetic energy, and penetration thickness, facilitating comprehensive trajectory optimization.The weighted summation method is employed to ascertain the optimal solution,with the simulated annealing-particle swarm optimization (SA-PSO)utilized to address the optimization challenge. The results demonstrate that this approach converges to the optimal solution more efficiently compared to traditional algorithms. Compared to the original design, the optimized solution increases landing kinetic energy by 109.97%, enhances penetration thickness by 75.11%, and the travel distance is shortened by 30.01%. The proposed method significantly improves overall ballistic performance. Moreover, this method circumvents the decline in other objectives that commonly arises during the optimization of a single target.

Aiming at the problem that the synthetic aperture radar (SAR) image detection model is difficult to balance the detection accuracy and model lightweight, this study proposes a lightweight SAR image target intelligent detection method based on YOLOv11 s. This method first replaces the backbone network with an efficient FasterNet structure, which significantly reduces the number of model parameters; secondly, the independently developed EMIBC module is innovatively integrated into the C3K2 module, which effectively improves the recognition ability of the model for small targets and multi-scale targets. Thirdly, the dynamic upsampling (DySample) is used to replace the traditional upsampling method to optimize the processing efficiency of the feature fusion stage. Finally, the Inner-SIoU loss function is introduced to replace the original CIoU bounding box loss, which further improves the training effect and feature extraction ability of the model. The experimental results on the HRSID dataset show that the improved model reduces the computational complexity index GFLOPs by 2.79 %, and the detection accuracy index mAP is increased by 7.35 %, which better realizes the balance optimization of model lightweight and detection accuracy.

In recognition of the intricate structure and multifaceted nature of military communication networks, two typical models of such networks are established, respectively, from the viewpoints of interdependent and multilayer networks. With a particular emphasis on interdependent networks, an analysis is conducted on the importance of nodes and the robustness of military communication networks through node importance analysis and cascading failure assessments. To validate the models, a modeling and simulation is performed based on the communication network organization plan outlined in a military wargame scenario. The results obtained encompass the ranking of node importance and the impact of various factors, such as attack methods, node capacity, and load distribution, on the cascading failure process within military communication networks. These findings offer valuable insights for the analysis of connectivity and robustness in military communication networks.

Addressing the challenge of quantifying combat capabilities in complex military scenarios, this study proposes a novel capability measurement method based on the potential outcomes framework. This approach first define the object to be evaluated as an intervention variable and selects observable metrics that directly reflect operational effectiveness as outcome variables. Next, a causal graph is constructed, encompassing the intervention, combat capability, and operational effectiveness, thereby explicitly establishing the causal pathways between abstract combat capabilities and measurable effectiveness indicators. Finally, through causal effect inference and statistical significance testing, we transform the problem of measuring combat capability into a function of effectiveness indicators. Case analysis demonstrates that the proposed method, by utilizing effectiveness indicators as a measurement tool, can achieve indirect quantification of combat capabilities, offering a valuable reference for solving the quantification problem of combat capabilities.

With the continuous development of ship information technology, automatic identification system (AIS) plays an increasingly important role in marine traffic management and ship navigation safety. In this paper, a ship identification method based on AIS data is proposed to solve the problem of military-civilian type identification of ships. The method establishes a sensitive sea area model for the distribution of key points of ship trajectories, calculates the probability of military-civilian attributes of ships, and constructs a category classifier to judge the military-civilian attributes of ships. The proposed method uses cheap, all-domain coverage and high frequency AIS information to identify the military and civilian types of ships, which can detect and warn the threatening behaviors existing in the sea area in advance, provide support and auxiliary decision-making for subsequent response and disposal.

In this paper, the use of airborne sonar under deep sea conditions is studied. The sonar detection distance model is established by using the sonar equation. Based on the analysis of the structure and variation characteristics of the deep-sea sound velocity profile, the underwater sound field distribution under different deep-sea sound velocity profile structures and different sonar detection depths is simulated by using the ray acoustic model. The submarine’s activity law is analyzed, and the relationship between the sonar working depth and the average detection distance is obtained. The results show that under deep sea conditions, the increase of the working depth of the airborne sonar is beneficial to the detection of unknown depth targets. The research results have guiding significance for guiding the use of aerial sonar in deep sea conditions.

In response to the problem of large computational complexity and false detections in target detection by drones. A target recognition method based on lightweight networks has been proposed. Based on the YOLOv5 object detection algorithm, the algorithm has been optimized using the FasterNet lightweight network architecture, which reduces the number of network parameters and improves the efficiency of the algorithm. In order to accurately capture and emphasize key information in the input sequence, and further enhance the performance of the algorithm, a parameter free attention mechanism SimAM is introduced. The results indicate that this method is an optimized application of object detection technology, which can better balance the relationship between detection speed and accuracy, and achieve better detection results in unmanned aerial vehicle aerial image detection tasks.

Against the backdrop of duty patrols, the paper focuses on addressing the issue of patrol prevention and control path planning in a key area of a city, taking into account multiple factors such as road distance, terrorist attacks and harassment, and weather conditions. A Matlab simulation is used to generate a city area map, based on which the urban patrol path is planned using an improved ant colony algorithm. Comparative experiments are conducted to verify the feasibility, reliability, and efficiency of this improved algorithm in patrol path planning. The experimental results show that the improved ant colony algorithm has fewer iterations and takes much less time than other methods. It is also suitable for complex patrol route planning problems and can provide favorable technical support for patrol prevention and control.

Aiming at the typical training demands of commander training for joint operations teaching in military academies, such as mass joint operations knowledge inquiry, battle case traceability and re-pushing, combat scene cognition and blue army confrontation, four novel commander training modes are put forward from the perspective of top-level design which are the theoretical teaching oriented commander training mode with question answering, battle case study oriented commander training mode with extension, scenario teaching oriented commander training mode with cognition and countermeasure training oriented commander training mode with confrontation. Then the essential technologies such as military knowledge intelligent answering based on large language model, parallel battlefield and situational cognition, construction and analysis of military event evolutionary graph and blue army behavior tree modeling integrating online learning are analyzed in detail which support the four novel training modes. This study can provide methodological support and implementation reference for military academies to carry out commander training with certain characteristics of digital-intelligence integration.