Military intelligent technology is the application of artificial intelligence technology in the military field. It includes basic technologies such as military generative intelligent technology and embodied intelligent technology, as well as core technologies mainly consisting of intelligent perception and cognitive technology, intelligent planning and decision-making technology, autonomous unmanned system technology, swarm intelligence technology, human-machine hybrid intelligence technology, and brain science technology. Currently, the development and application of military intelligent technology are still in the frontier exploration phase. This paper mainly delves deeply into the major impacts of the mature development of future military intelligent technology on army offensive battles. Through a meticulous analysis of multiple crucial dimensions, including the unmanned autonomy of offensive forces, the multi-domain integration of battle areas, the rapid scaling and conversion of time and space, the human-machine collaboration in battle command, the autonomous linkage of attack actions, and the focus on intelligence control as the battle center of gravity, it endeavors to clearly illustrate how military intelligent technology fundamentally transforms the existing forms of army offensive battles, offering a reference for relevant theoretical research and practical exploration of future army offensive battles.

Aiming at the problems of high energy consumption and weak anti-interference ability caused by insufficient regulation accuracy of topological state in UAV swarm cooperative control, a UAV swarm cooperative control method based on topological phase transition theory is proposed. First, the topological phase transition theory is mapped to the field of UAV swarm control, and a UAV swarm cooperative control model based on topological phase transition theory is constructed. Second, a closed-loop control method based on phase transition perception, phase transition judgment, phase transition regulation and phase transition feedback is proposed, a strategy for dynamically adjusting the communication radius is designed, and the critical communication radius is determined through random graph theory derivation and simulation. Finally, the feasibility of the method is verified by simulation. The results of multiple simulation experiments show that, compared with the UAV swarm phase transition control method based on the self-propulsion mechanism of bird flocks, the anti-interference ability of the proposed method is improved by 40%; compared with the control method based on reinforcement learning, the energy consumption is reduced by 5%, and the computational complexity is significantly reduced.

To improve the command and control effectiveness of manned-unmanned cooperative combat, this paper conducts a systematic study based on the network information system. Firstly, the application scenarios are analyzed from three typical dimensions: battlefield reconnaissance and surveillance, fire strike and interception, and battlefield support and sustainment; the cooperation requirements and effectiveness characteristics under different combat scenarios are elaborated. Secondly, three novel command and control modes are summarized, namely the flat network mode, intelligent autonomous mode and virtual immersive mode. On this basis, a three-layer system architecture is constructed, consisting of the network information system layer, command and control decision-making layer and manned-unmanned cooperation layer, with the functional positioning and interaction relationships of each layer clearly defined. Furthermore, four key technologies are deeply analyzed, including distributed combat cloud, intelligent command and control decision-making, dynamic mission planning, and augmented reality presentation. Finally, suggestions for future development are put forward from three dimensions: combat concept innovation, system construction path and technology research and development orientation.

The command decision-making style of military commanders significantly impacts combat outcomes, yet existing research lacks objective quantitative analysis methods. To address this gap, this study establishes a quantitative index system through three behavioral dimensions derived from wargame replay data: terrain utilization, weapon deployment, and operational coordination. Utilizing hierarchical clustering algorithms, we achieve unsupervised classification of command decision-making styles. Three distinct tactical patterns are identified: terrain and firepower-dominant style, coordinated assault-efficient style, and balanced and steady style, each exhibiting marked disparities in tactical preferences. Corresponding counterstrategies are systematically proposed for each identified style. This research provides theoretical support for optimizing command decisions and developing wargame AI systems, demonstrating substantial potential for advancing intelligent military decision-making through data-driven style characterization.

In modern offensive and defensive operations, seizing air superiority is a decisive factor influencing the war situation. To achieve this, the Red Force must accurately predict the maneuver strategies of the Blue Force’s air defense fire units and formulate efficient joint reconnaissance-strike integrated operations. From the perspective of the Red Force, this paper constructs a multi-objective optimization model for the maneuver of the Blue Force’s air defense fire units under multiple constraints such as the air defense fire unit-target allocation and position deployment, to maximize the overall defensive capabilities and minimize the total maneuver time of air defense fire units. The ideal point method was employed to transform the multi-objective problem into a single-objective form, while equivalent simplification was applied to the nonlinear constraints in the model, thereby designing an efficient corresponding solution algorithm. With the aid of LINGO software for implementation, the Pareto optimal solution set under different weight combinations was ultimately obtained. Simulation results show that the weight coefficients determine the trade-off relationship of operational effectiveness between defense effectiveness and maneuver efficiency. The proposed model provides commanders with decision-making schemes oriented to different operational intents, enabling them to make scientific and rational decisions based on the battlefield situations.

In order to improve the positioning accuracy of Underwater Wireless Sensor Network (UWSN) nodes, a zone-partitioning-based localization (ZPL) algorithm is proposed based on triangle geometry theory. The algorithm uses the triangular area formed between two anchor nodes and the node to be located, takes the ranging result as the side length, and iteratively refines the area according to the geometric theory to achieve position estimation. In the process of zone partitioning iterative positioning, linear algebra operations with low complexity are mainly used to improve the computational efficiency. After simulation comparison and analysis, the ZPL algorithm has more stable performance than the traditional positioning method, effectively increases the number of locatable nodes, and reduces the positioning error.

The fundamental principles and technical advantages/disadvantages of Deep Packet Inspection(DPI) and Deep Flow Inspection(DFI) technologies are primarily elaborated, it analyzes the current management status of IPTV systems in existing Service networks, through collecting and analyzing traffic characteristics of IPTV service control information and video data, we establish a critical video feature database and priority hierarchy based on traffic patterns, leveraging DPI/DFI technologies enables precise identification of various IPTV-related traffic types, a refined IPTV traffic management process is designed, implementing an integrated hardware-software IPTV traffic management system, a testing environment was constructed with evaluation metrics to examine: IPTV video display under normal bandwidth conditions, video performance during bandwidth threshold exceedance and video quality after implementing priority protection strategies, experimental results demonstrate the DPI/DFI based IPTV traffic control achieves accurate traffic identification and granular control, effectively mitigating the impact of network congestion on service quality.

The intelligent upgrade of network-centric operation and maintenance systems relies on the efficient deployment of large language models (LLMs). However, resource-constrained edge environments and dynamic adversarial threats pose dual challenges to traditional fine-tuning methods. Existing approaches often address computational efficiency and model security in isolation, leading to lightweight strategies that compromise defense capabilities or security mechanisms that exacerbate resource consumption. To resolve this conflict, this paper proposes a Lightweight and Secure Co-Tuning Framework (LST-Framework) that achieves multi-objective optimization through dynamic parameter sparsification and hierarchical defense mechanisms. The method first constructs a task-sensitive dynamic parameter activation strategy, identifying redundant weights via spectral analysis of the Hessian matrix to establish sparse fine-tuning paths, reducing computational complexity to 23% of full-parameter fine-tuning. Simultaneously, a three-tier defense system—comprising adversarial input detection, lightweight gradient masking, and output uncertainty verification—enhances robustness against white-box attacks with only an 8% increase in inference overhead. Experiments on typical network operation datasets (e.g., CIC-IDS2017) demonstrate that the LST framework maintains 98.3% original task accuracy while compressing memory usage by 51.7%, reducing inference latency by 32.4%, and successfully resisting 94.6% of PGD-targeted attacks. This work provides theoretical and technical foundations for deploying large models in low-resource, high-security scenarios such as 5G core networks and industrial IoT.

In order to achieve dynamic management of IoT access permissions and improve the security of IoT data, a fine-grained control method for IoT terminal access considering permission thresholds is proposed. Firstly, the AES encryption algorithm with time stamps is used to encrypt sensitive data in the Internet of Things, improving the security of sensitive data. Secondly, loyalty is introduced to calculate the trust value of IoT visitors, and the identity of visitors is determined by comparing the trust value with the permission threshold, and data access permissions are assigned to visitors. Finally, the priority of non malicious visitors is determined through fuzzy comprehensive evaluation method to achieve fine-grained control of IoT terminal access. The experimental results show that this method has low data encryption cost and good access control effect.

To address the challenges of rapid rumor dissemination and the tendency of traditional algorithms to fall into local optima in social networks, this paper proposes a real-time rumor refutation method based on a hybrid Shuffled Frog Leaping Algorithm and Particle Swarm Optimization (SFLA-PSO). This approach integrates the clustering cooperative local search capability of SFLA with the fast global convergence advantage of PSO. Firstly, a rumor dissemination model incorporating users' social ties and trust degrees is constructed. Users are categorized into different age groups, and differentiated modeling strategies (SFLA or PSO) are applied according to their cognitive characteristics. Secondly, the SFLA-PSO hybrid optimization algorithm is designed to dynamically update user trust levels through local deep search and global information interaction, thereby efficiently blocking rumor propagation. Simulation results demonstrate that, compared with standalone SFLA and PSO algorithms, the proposed hybrid algorithm exhibits significant advantages in convergence speed and computational efficiency. It can guide user cognition towards the truth more rapidly and comprehensively, effectively curbing the spread of rumors.

Weapon system combat simulation models are core resources for equipment digitalization and urgently require standardized construction. This study analyzes the four main approaches to military simulation model development both domestically and internationally, identifies the three-layer dependency of simulation models on system simulation platforms, proposes a problem-oriented hierarchical methodology for model systems, and designs a layered framework for cross-platform modeling of weapon system combat simulation across three levels: model specifications, development frameworks, and model architectures. A reference framework supporting hierarchical construction of weapon system combat simulation models is provided, which can serve as a comprehensive technical solution for the development of combat simulation models in the weapon system domain.

In response to the coexistence of structured business data and unstructured PDF text data in the equipment domain, existing information extraction methods are costly and struggle to adapt to the diversity and dynamic nature of equipment-related data. This study investigates a PDF data extraction method for the equipment domain based on the DeepSeek large language model. The paper constructs an information extraction framework encompassing attributes such as basic properties, technical parameters, performance metrics, and timeliness indicators. A dataset extracted from equipment technical manuals, maintenance procedures, and acceptance standard documents is designed. By creating zero-shot prompt templates and conducting experiments with DeepSeek models of varying scales, the effectiveness of this method in constructing equipment knowledge graphs is validated. The experimental results demonstrate that the pre-training capabilities and fine-tuning techniques of the DeepSeek large model significantly improve the accuracy and recall rate of PDF text information extraction in the equipment domain. This provides an efficient and flexible solution for the intelligent management of equipment data.

Infrared characteristic simulation technology has been widely applied in multiple aspects such as scheme design optimization, risk assessment, predictive analysis, and performance testing. This paper classifies the infrared characteristic simulation technologies into five categories, and analyzes the infrared simulation technologies based on experimental data, empirical models, physical models, measured images, and deep learning from the aspects of basic principles, development history, analysis steps, and technical advantages and disadvantages. Combined with the practical requirements such as improving simulation accuracy, adaptability to complex scenarios, and real-time simulation performance. The future development trends of infrared characteristic simulation technology are further prospected from multiple aspects such as the integration of physical models and data-driven, high-precision modeling, multi-physical field coupling, and hardware guarantee, providing references for research and development in related fields.

In Python, mathematical and plotting libraries were utilized to conduct a refined modeling of the submarine-attack process by homing aerial depth charges. Under specific conditions, simulation analyses were performed on hit probability to evaluate the weapon system’s anti-submarine effectiveness. The study focused on multiple key parameters (using assumed values), including sinking speed, effective search radius, search sector angle, initial search depth, and dispersion error. Through simulation experiments, the impact of these parameter variations on hit probability was investigated, leading to a comprehensive assessment of combat performance. The simulation results demonstrate the objectivity and credibility of the analysis while highlighting Python’s practicality and versatility in solving such problems.

To address the problem of battlefield airspace utilization effectiveness evaluation, considering the particularity of battlefield airspace operation, an evaluation index system consisting of 4 dimensions and 18 elements is established in accordance with the principles of safety, effectiveness and flexibility. On this basis, a combined weighting method based on the queuing rank concept is adopted to determine the indicator weights, and the corresponding relationship between the normal cloud model and statistical concepts such as random variables and probability measure spaces is established, the cloud model is then applied to conduct the evaluation of battlefield airspace operational effectiveness. Finally, case verification and comparative analysis with the fuzzy comprehensive evaluation method and the fuzzy analytic network process (FANP) demonstrate that the evaluation results of this method are more reliable,and the method integrates both fuzziness and randomness, and can provide a reference for improving the utilization effectiveness of battlefield airspace.

In order to test the current level of aerospace equipment support training ability, an evaluation based on the “AHP+cloud” model is carried out. First, the necessity and value of conducting the evaluation are analyzed, laying a preliminary theoretical foundation for the evaluation. Secondly, the index system is constructed according to the steps of clarifying the reference basis, standardizing the construction actions, and analyzing the index factors, and the “AHP+cloud model” method is designed, and an example verification is carried out, obtaining objective and specific conclusions of “good”. Finally, targeted opinions and suggestions are put forward for the identified problems, so as to better serve the aerospace industry.

With the continuous development of deep learning in driving behavior analysis, traditional methods have shown limitations in processing massive video and image data.To address the problem of driver distraction recognition, a distracted driving recognition method based on keyframe guidance and adaptive channel MAMBA is proposed.It includes:(1) guiding the model to focus on discriminative information through a keyframe guidance mechanism;(2) enhancing spatial feature representation by integrating edge feature enhancement and channel attention mechanism;(3) capturing the dynamic evolution of driving actions via multi-frame temporal modeling. Experimental results demonstrate that the proposed method achieves 99% accuracy on the SAM-DD dataset,and improves the accuracy by 7 percentage points compared with VideoCLIP in the cross-dataset transfer task (SynDD1->StateFarm). This study provides a high-precision and robust solution for distracted driving recognition,and offers a reference for the optimization of temporal modeling and modular design.

Traditional dynamic scheduling methods for rail transit signals lack real-time dynamic optimization capabilities and exhibit rigid scheduling strategies, leading to prolonged average vehicle waiting times and increased average queue lengths after scheduling. To address this issue, this study designs a spatiotemporal evolution modeling and dynamic scheduling method for rail transit signals. The spatiotemporal evolution modeling provides accurate and real-time input data for dynamic scheduling of rail transit signals. A dueling double deep Q-network (D3QN) is designed by integrating double deep Q-networks (DDQN) and dueling deep Q-networks, leveraging decoupling and dueling structures to achieve dynamic scheduling of rail transit signals. Case study results demonstrate that the proposed method effectively enables dynamic scheduling of rail transit signals at experimental transportation hubs. The average vehicle waiting time using the proposed method remains below 250 seconds, and the average queue length is consistently around 20 meters, indicating high operational efficiency of the rail transit system under dynamic scheduling.

In the era of intelligent warfare, data has transformed into a critical strategic resource that determines military competitive advantage. As a pioneer in global military data capability development, the United States’ strategic evolution offers a typical case study. Based on core strategic documents and recent practices from the U.S. Department of Defense and its service branches, and building upon a review of existing research, this paper proposes a three-stage evolutionary logic for the U.S. military’s data strategy: Network-Centric, Data-Centric, and Decision-Centric. The study finds that the evolution of the U.S. military’s data strategy demonstrates a clear logical progression: from connecting everything in the Network-Centric stage, to governing assets in the Data-Centric stage, and ultimately moving towards empowering victory in the Decision-Centric stage. This paper systematically analyzes the strategic thinking, key initiatives, and internal driving forces of each stage, and details the differentiated practical paths of the various service branches. Finally, by considering the realities of China’s military data development, it offers targeted implications from the perspectives of top-level design, implementation systems, talent strategy, development models, and security governance, aiming to provide theoretical reference for building a military data capability system that meets the demands of future warfare.

Based on open source information on the Internet, the development planning of U.S. unmanned submersible vehicles are sorted out. The equipment construction status quo is analyzed by classifying four categories: super-large, large, medium and small. And at the same time, according to operational needs and technological development status, eight typical military application scenarios of U.S. strategic deterrence, battlefield reconnaissance, anti-mine operations, target strike, payload transportation, network nodes,teamfight with no one and cluster combat are discussed. Based on further analyzing experience and teaching of the U.S. military UUV development planning, equipment construction, military application, the five development strategies of the U.S. military unmanned submersible vehicle cascade, systematization, generalization, standardization and automation are summarized.

To sustain its global maritime dominance, the U.S. Navy is actively advancing the military implementation and intelligent system design of AI. Through a systematic examination of 146 AI-related subprojects funded by the U.S. Navy for FY2025, categorized under seven fundamental AI technological frameworks, this analysis reveals that the U.S. Navy is deploying single or integrated AI models across four critical operational domains: information management, command-and-control systems, logistics operations, and combat training. Among these, four principal models demonstrate predominant utilization: hyper-personalization, recognition, predictive analytics & decisions, autonomous system model. Three strategic recommendations are proposed: firstly, leverage AI-enhanced multispectral sensing systems to achieve comprehensive battlefield awareness; secondly, establish a human-machine cognitive symbiosis framework to revolutionize command-and-control paradigms; thirdly, develop an intelligent logistics ecosystem featuring demand forecasting algorithms and dynamic resource allocation mechanisms. These proposals collectively provide a multidimensional perspective on the U.S. Navy’s current AI integration landscape.