Aimming at the multitudinous battlefield situation information and the difficulty in recognizing the changing trends, based on large models, a situation awareness agent and an intelligent situation awareness inductive method are proposed. Starting from cognitive concepts and combining the abstractness and embodiment characteristics of agents, three key components in the construction of agents have been clarified: learning environment, memory mode, and knowledge generation mechanism. The architecture of the battlefield situation awareness agent is designed, including memory component, planning component, execution component, evaluation component, and key points for agent training. In the long-term memory component, based on the modeling characteristics of complex battlefield states, the paper discusses the application of large language models, multimodal large models and large sequence models.

With the development of new weapon systems and the application of intelligent technology in military fields, the research of intelligent situation awareness technology has become the hot spot among military powers. Firstly, the concepts of situation awareness and intelligent situation awareness are introduced. Secondly, In view of the characteristics of modern warfare, five critical demands of intelligent situation awareness are analyzed, including intelligent combat system of system analysis, intelligent awareness of combat window, combat situation evolution and prediction, intelligent cognition production representation, and intelligent combat situation retrospective. With domestic and foreign research literature, the key technologies and countermeasures involved are analyzed. The relevant results have certain reference significance for the future development and technological breakthroughs of intelligent situation awareness.

To study the intelligent decision making methods under limited decision samples, aiming at the problems that operational decision-making experience is difficult to express and the training samples for intelligent decision learning are limited, based on the joint operational simulation and drill environment, a decision-making method based on generative adversarial imitation learning is proposed. This method integrates the operational decision-making experience representation and learning process. On the basis of high-level decision-making and low-level action, rule definitions are used to specify the logic of task execution, and generative adversarial imitation learning algorithms are utilized to improve the generalization ability of intelligent agents in scenarios. This method achieved expected results in the constructed typical adversarial scenarios. The algorithm training converged and the decisions output by the intelligent agent are reasonable. Preliminary experimental results indicate that generative adversarial imitation learning, as an intelligent operational decision-making method, has value for further research.

Intelligent command and control has become a new style of modern warfare command, and the high dynamic, strong adversarial, and massive data battlefield environment has put forward higher requirements for command efficiency and task execution ability. The traditional human-computer interaction mode has low interaction efficiency and lacks a human-centered interaction design, which cannot fully utilize the effectiveness of underwater command and control. This article is based on intelligent human-machine interaction and principal component analysis to design an interaction module that can generate an adaptive interface for underwater command and control based on user and business needs. By building an underwater command and control intelligent simulation platform, it has been confirmed that adaptive intelligent human-machine interaction can simplify the command and control process, reduce interaction time, and improve combat efficiency, with strong application value.

Aiming at the difficulties in the quantitative verification and evaluation of underwater countermeasures for our underwater platform and the lack of a guided maneuvering evasive combat model, a prediction model of underwater countermeasures based on big data learning is designed. Firstly, the underwater countermeasure modeling of the underwater platform is carried out, and the large data set of avoidance probability is obtained by several rounds of simulation based on Monte Carlo method. At the same time, in order to solve the problem of poor time efficiency under massive simulation, BP neural network prediction algorithm is proposed for big data learning to provide accurate, fast and visual antagonistic results. The test results show that under the test environment set in this paper, the average prediction error of BP neural network prediction algorithm is 7.28% respectively, which can effectively predict the avoidance probability of the underwater platform and provide data support for the commander’s command decision.

The breakthrough and progress of intelligent gaming technology with deep reinforcement learning as the core in the field of games provide a method reference for the research of agents in sea-air wargames. The architecture design of the agent is the primary core key problem that needs to be solved, and a good architecture can reduce the complexity and difficulty of training and accelerate the convergence of policies. A stochastic game model of sea-air cross-domain cooperative decision-making has been proposed, and its corresponding equilibrium solution concepts have been analyzed. Based on the analysis of typical agent frameworks, aiming at the decision-making gaming process of sea-air wargames, and then an agent bi-level architecture based on multi-Agent hierarchical reinforcement learning is proposed, which can effectively solve the problems of collaboration and dimensional disaster. The key technologies are analyzed from four aspects: force coordination, agent network design, adversary modeling and training mechanism. Hoping to provide architectural guidance for the subsequent design and implementation of sea-air wargaming agents.

As the key stage of operational concept development, capability requirement analysis is the process of analyzing solutions to specific operational problems, and defining various demand elements such as capability indicators, indicator thresholds, and indicator relationships. This analysis process needs to be supported by a scientific and reasonable indicator system with military significance. This indicator system should not only have the traditional purpose, integrity and measurability of indicators, but also have the characteristics of integration and hierarchy, emergence and aggregation, and qualitative and quantitative combination. Aiming at the above requirements, it takes the operational concept of "stereoscopic projection" as the research object. Firstly, based on DoDAF2.0, an operational concept architecture model with capability requirements as the core is built; Based on the characteristics of capacity demand analysis, a super network indicator system including three node networks is designed; Finally, according to the designed super network structure, based on the built operational concept architecture model, the capability requirement analysis indicator system is established, and the effectiveness of the method is analyzed.

To solve the problems of unstructured data organization, insufficient detailed description of entity relationship and lack of standardized expression of knowledge representation in the construction of knowledge graph in the field of early warning equipment. Based on the Protégé ontology construction tool, this paper divides the early warning equipment system from top to bottom into equipment principle knowledge class, management support resource class, equipment operation and application class, etc., and then refines them. So as to define the concept class and divide the hierarchical relationship, and extract the entities, attributes, relationships and other knowledge units in specific equipment information. Then, the domain knowledge of early warning equipment is modeled, and a scientific and complete knowledge representation framework of fine-grained early warning equipment is explored. Taking “long-range early warning phased array radar” as an example, ontology instance filling and visualization are carried out. Finally, the ontology of early warning equipment is formally represented based on ontology description language. The ontology lays a semantic foundation for the construction of domain knowledge graph based on multi-source high-quality data, and has important practical guiding value for realizing intelligent management support and operational application of equipment such as equipment capability portrait, development trend analysis, equipment fault diagnosis, and health status management under the background of big data.

Starting with the concept, principle and classification of AI technology, this article focuses on the analysis of three main scenarios and eight key directions of AI military applications. It comprehensively uses the maturity evaluation method of technology patent analysis and the Gartner technology maturity curve to evaluate the maturity of eight key directions of AI military applications.

A density clustering-based health monitoring model for complex equipment is proposed for the complex equipment historical data which often has the characteristics of non-spherical shape. The local density and inter-class distance of each sample are estimated from the historical data, and the statistical properties of both are considered to determine the clustering center of the data. For the newly collected complex equipment health status monitoring data, if it is reachable with the density of the clustering center, the complex equipment is considered to be in a healthy state, otherwise it is in a non-healthy state. An actual complex equipment data set is analyzed by numerical simulation techniques, as well as visualization techniques such as scatter plots, box plots and parallel coordinate systems are used to verify the reliability of the calculated results, and the simulation results show that the proposed method can effectively monitor the health status of complex equipment.

The evaluation of equipment comprehensive support effectiveness is affected by research and development support, personnel support, equipment support, maintenance support, and other aspects. How to evaluate the equipment’s comprehensive support effectiveness scientifically, objectively, and reasonably plays an important role in improving the equipment’s comprehensive support effectiveness of the army and promoting the generation of combat effectiveness of the army. Based on the research and analysis of related issues, the evaluation index system of equipment comprehensive support effectiveness is constructed. The analytic hierarchy process (AHP) and fuzzy comprehensive evaluation are combined to determine the weight of each evaluation index factor and the reasonable evaluation of support effectiveness. Finally, the equipment’s comprehensive support effectiveness is evaluated through the example analysis system. The results show that this method has certain reference significance and guidance for scientific and reasonable evaluation of equipment comprehensive support effectiveness.

The research emphasized the invulnerability of tree command network and Cascade Command network based on the normal duty command model. Random attack and deliberate attack are employed on three kinds of duty command models. The paper found out the best anti-destruction performance model by comparing and analyzing indicators like the indexes of maximum connectivity subgraph size, network efficiency, global aggregation coefficient etc. The collective influence algorithm is used to identify the key nodes of the command network with the best anti-destruction performance, and compared with the traditional key node identification method based on node degree, it is proved that it is more accurate to identify key nodes based on the collective influence algorithm which could provide theoretical guidance for the optimization and protection of the future regular duty command network.

From the perspective of the historical development of the command of maritime operations, it is found that the flow of command information gradually increases, the flow direction gradually increases, and the flow velocity gradually increases. Under the influence of the change of the command information flow, the command structure of maritime operations has undergone a process of change from simple to complex, and its internal structure is constantly optimized, and the external structure is constantly expanded and the operation mode is constantly sensitive.

In order to scientifically evaluate the command capability of a joint combat command organization, the paper establishes a command capability indicator system for the joint combat command organization, analyzes the interrelationships between indicators at various levels and the interrelationships between elements within the indicator group, and uses Super Decisions software to establish a relationship diagram of command capability evaluation indicators for the joint combat command organization, obtaining the weight of the final indicator system, collects two groups of command organization exercise evaluation data and calculates the evaluation results of capability. By analyzing the weight of indicators, it is pointed out that to improve the command capability of joint operations organizations, it is particularly important to focus on three key points: planning and decision-making capabilities, intelligence reconnaissance capabilities, and operational control capabilities.

An active disturbance suppression and model predictive control (MPC) strategy is proposed for the quadrotor precision trajectory tracking control problem under disturbance action and model uncertainty. The model predictive controller estimates and compensates the disturbance by the extended state observer (ESO) and disturbance observer (DO) to achieve the position loop accurate control. In the presence of external disturbances and parameter uncertainties, it is demonstrated through simulation experiments that the proposed method improves the robustness to modeling errors and disturbances, while achieving smooth tracking of the reference trajectory.

Aiming at the problems of difficult implementation and poor robustness of quadrotor UAV attitude control in unknown environment, an intelligent attitude control method based on deep deterministic strategy (DDPG) algorithm is proposed. Firstly, based on the Euler-Poincare equation, the dynamic model of the quadrotor is established by computer symbolic derivation. Secondly, the attitude controller of the quadcopter is designed based on the DDPG algorithm, and the attitude error, attitude angular velocity error and control law penalty items are introduced into the reward function design. Finally, the performance of the controller is verified by setting different initial state values, changing the structural parameters of the quadcopter, and introducing noise. The simulation results show that the controller can guide the quadrotor to respond quickly to the desired attitude and maintain stability, and also show good generalization ability.

Frequency Diverse Array (FDA) is a brand-new electronically scanned radar system. The time-range-angle three-dimensional correlation characteristics of its beampattern have broad application prospects in the fields of electronic countermeasures and security communications. One-dimensional uniform linear FDA array (ULFDA) has the problem of angle-range coupling in the transmit beampattern. To solve this problem, this paper adopts sine based on the symmetrical subarray density taper array structure. The frequency offset obtains the dot-shaped beampattern pointing control method at the target point position. In addition, we analyze the time-varying problem of the FDA pattern, and get the conclusion that the mainlobe of the density tapered array sin-FDA based on the subarray structure has a fixed direction in the angle dimension and changes in the range dimension.

As a new type of intelligent mine, homing mine plays an important role in modern maritime warfare. In order to improve the interception probability of the passing target by the mine barrage composed of homing mine. Firstly, in order to obtain the detection area of homing mine in the battlefield environment, the target detection system of homing mine was modelled in this dissertation. Then, based on the genetic algorithm, the deployment position of the homing mine is optimized. By simulating the situation when multiple targets pass through mine barrage before and after optimization, obtain the average probability of the target triggering the homing mine before and after the optimization. The superiority of homing mine placement after optimization is verified. It has guiding significance for the laying of mine positions of homing mine.

The torpedo discovery probability is associated with launch situation before homing torpedo launched. In order to meet the need, a method for predicting the tendency probability change is proposed. The torpedo launch situation prediction model is built according to the current attack situation,and the torpedo launch situation sequence is generated. Based on the principle of normal distribution, target dispersion is processed. The area ratio method is used to calculate the single launch point discovery probability of acoustic homing torpedo, and length ratio method is used for wake homing torpedo. Based on the above processing methods, the continuous discovery probability of homing torpedo at launch time is calculated, and the trend chart is generated. Finally, the prediction result of finding probability variation trend of homing torpedo is tested and verified by using typical launch situation.

Aiming at the problem of uncertainty and complexity of battlefield situation, which leads to the poor accuracy of risk prediction in the process of research and judgment, it is proposed to build a situation-risk model and construct the risk possibility distribution in order to realize the accurate prediction of risk. Firstly, the risk prediction index system is established from the enemy situation, our situation and the battlefield environment situation research and judgment content; secondly, the role of situation information on the risk is analyzed, and the risk model such as exponential, Z-type, tangent and so on is constructed and the risk value interval is calculated; then, based on the fuzzy function constructing method in the possibility theory, and relying on the weighting difference and uncertainty measurement, the risk possibility distribution is constructed within the risk value interval. Finally, an example of a defense mission is used to verify the feasibility and reasonableness of the proposed method by combining the expert system and cloud modeling methods in an experiment. The results show that the proposed method can fully take into account the uncertainty of situational information in risk prediction, which provides an important basis for subsequent planning decision-making.

In response to the existing gun sound recognition and positioning tasks, which require separate identification and positioning, resulting in time-consuming computation, system redundancy, and complex development processes, this paper proposes to use a two-stage CRNN deep learning network model to complete the gun sound recognition and positioning tasks. Firstly, perform a logarithmic Mel transform on the collected gunshot signal and calculate the generalized phase transition cross correlation spectrum as input to the network model. Secondly, in the first stage, the gunshot signal is identified through the CRNN network. Finally, in the second stage, the introduction of a mask is used to determine whether the CRNN network weight sharing is implemented for localization. The method proposed in this article can effectively solve the problems of separate recognition and positioning tasks, system redundancy, and complex development processes in traditional methods, and has certain application value in achieving joint recognition and positioning.

Aiming at the configuration problem of amphibious assault ship formation in submarine defense, a circle model of submarine torpedo limit range is established from the analysis of submarine threat to formation. Through model analysis and simulation calculation, it is concluded that amphibious assault ships should be positioned in the later position of threat circle center, and the optimal position calculation model is given. Based on the model of submarine threat range and threat range and the model of submarine defense range of warning ships, the optimal allocation method of escort forces is given, which provides a reference for the submarine defense of amphibious assault ship formation at sea.