中国科技核心期刊
中国指挥与控制学会会刊
军事装备类重点期刊
中国指挥与控制学会会刊
军事装备类重点期刊
中国指挥与控制学会会刊
军事装备类重点期刊
This paper reviews the CARACaS autonomy framework, its core components, the related maritime demonstrations, some applications and newly developments. Firstly, it concisely describes the relationships of CARACaS’s sub-systems, along with the principles of its software architecture, the characteristics of its behavior engine, dynamic planning engine, perception engine and world model. Secondly, the typical application of the CARACaS’s techniques in the Swarm Ⅱ harbor patrol demonstration is analyzed in detail, and discussed the evolution of the CARACaS’s perception system used in the USVs’ autonomous navigation. Thirdly, the Spatial Integrated Systems(SIS), as an industry leader in the development and integration of solutions incorporating next generation digital 3D data capture and processing technologies, has applied the CARACaS’s techniques to serve as the heart of the SIS’s autonomous capabilities, its products and systems of unmanned technologies are recommended. Finally, it also sums up some other unmanned system developments such as US Navy’s unmanned shipbuilding plans, unmanned tactical control systems, and it conclude the paper with some inspirations.
Information correlation (IC) is incompatible with target localization for distributed passive sensor (DPS) so that the cost function cannot be determined and the modeling is difficult when the number of targets is unknown. So a feasible correlation matrix(FCM) is defined for IC. Cost function based on maximum likelihood probability distribution is determined. Optimal model for IC between 2 sensors is put forward about FCM. The extremum theorem of cost function is proved. Simulation results indicate the efficiency of the proposed model for IC, which may have reference to information fusion system for DPS.
In practical applications, the integrated navigation system has the influence of delay in signal transmission and calculation. Directly using the measurement data with delay will cause the accuracy of the filtering algorithm to decrease or even diverge. In order to solve the filtering problem of this kind of measurement with random delay, this paper proposes the random delay Kalman filter algorithm. The core of the algorithm is to use new measurements to update multiple states in the past to compensate for the delay. The simulation results of the integrated navigation system of GNSS and INS show that the filtering algorithm designed in this paper can reduce the spike phenomenon after filtering and reduce the error of the estimation result when the measurement has a random delay.
Variational autoencoder (VAE) infers random missing data reduction is an emerging data filling method in recent years, but the traditional VAE algorithm has the problems of single posterior distribution and slow convergence of the evidence lower bound of variational inference. A new Bayesian inference network framework, Index-GMVAE, based on Gaussian mixture posterior distribution and missing data location mask to infer missing data restoration is proposed in this paper. The experiments are performed on the Mnist handwritten digit dataset and the Adult dataset. The simulation results show that the proposed algorithm can significantly improve the convergence speed of the evidence lower bound. Compared with the VAE algorithm, the accuracy of inferring missing data restoration is increased by about 10%. And it can effectively improve the inference performance of the probability hidden layer of model on both the image dataset and the heterogeneous binary classification dataset, which reflects the universality of data with different structures.
In this paper, the divergence problem of Taylor series method in target motion analysis is simulated and analyzed. Aiming at the singularity of the original method, an improved weight method is proposed. According to the results of simulation analysis, the selection principle of the iterative initial value is given. Finally, a “no danger of anything going wrong” divergent remedy scheme is given. The simulation results show that the improved method has no divergence for the common situation and the preferred initial value, but all converges, so it can be used in engineering practice “without worry”.
For the visual task accuracy of the target detection, the target detection is significantly decreased, and a end-to-end image density network is proposed to directly generate a clear image and do not require a pair of fogging images and real images. Integrating attention module into the generator network, considering the effects of different channels and different regions in fog images, and combining channel attention and pixel attention, a new feature focus module is proposed to give important Feature channels and more weights in pixel regions. Considering the different effects of high and low frequency information in the fog diagram, the frequency information is fused into the discriminator. Finally, the loss function is optimized, adding structural losses to ensure cycle consistency, decomposing the high and low frequency loss of the apparatus, reducing color distortion and artifact. Experiments show that this method achieves excellent performance on the image of public synthetic data sets and real-world, with better subjective visual effects.
Aiming at the problem of insufficient feature extraction of radar trajectory data, the accuracy of air target classification and recognition is low. This paper presents a method for extracting high-dimensional feature matrix of track data. The article first analyzes the trajectory data from maneuverability, cruising, flight area and high-level features, then statistical data features in different dimensions, extracting multi-dimensional trajectory data feature parameters, and finally forming a high-dimensional feature matrix of trajectory data. Experiments with measured track data show that the feature extraction is sufficient. The statistical average of the recognition rate verified by multiple machine learning methods is 92.4%, which proves the feasibility and stability of the algorithm in this paper. This method can be used as an effective means to improve the accuracy of track target recognition.
In order to study the influence of grid type and grid scale changes on the discrete error in the numerical simulation, CFD simulation is performed on the NACA2412 three-dimensional airfoil at 0°, 4°, 8° and 12° angles of attack, and the lift coefficient and drag coefficient under corresponding conditions are obtained. The Richardson interpolation method is used to study the spatial discretization errors of multiple sets of continuously changing structured grids and unstructured grids. The grid convergence index (GCI) method is used to analyze the error. Finally, the modified exact solution of the corresponding grid research group is obtained by grid convergence analysis. The results show that the discrete errors of the structured grid model are significantly less than those of the unstructured grid under the conditions of various angles of attack. The change of the mesh type has a greater impact on the lift coefficient, and the fine mesh model has better mesh convergence.
Assessment is an important part of the joint operations concept plan simulation experiment, the assessment framework is a kind of understanding and guidance for simulation experiment assessment activities. Through the study on assessment framework for simulation experiment of joint operational concept plan, further standardize joint operations concept plan assessment and analysis activities, from the perspective of evaluation analysis, deepen the understanding of joint operations concept plan simulation activities, provide more scientific and effective theories and methods for organizing and implementing simulation experiment activities and constructing and developing simulation experiment system.
Combat simulation can provide a series of valid means and methods for studying the complicate combat problems, such as combat tasks, combat process, combat operation, equipment effectiveness and system capacity. However, the combat simulation knowledge involves multiple domains, multiple disciplines, multiple levels and multiple military services, which leads to some problems, such as heterogeneous expressions, hard common understanding and lacks of reuse. This restricts the system development, action execution, efficiency elaboration of combat simulation. Providing the sharable, understandable, computable combat simulation knowledge model shall improve the implementing efficiency and applicable effectiveness. Hence, by combining the ontology engineering and combat simulation features, this paper studies the necessities of the combat simulation ontology, and proposes its construction principles and process, which will help construct the specific knowledge model.
In the field of maritime simulation training, the fidelity and antagonism of force behavior in battlefield simulation directly affect the training effect, and force behavior modeling technology is the core of maritime simulation. With the gradual deepening of deep reinforcement learning in military field, it is a feasible technical way to realize maritime force behavior modeling by combining rule reasoning and reinforcement learning it provides technical support for the construction of intelligent opponent in simulation training.
Automatic Identification System (AIS) data contains time series information related to ship traffic. Similarity search based on AIS data segment has also become a research hotspot in the field of intelligent shipping. AIS data belong to multivariate time series data. The traditional multidimensional data series similarity search methods reduce the dimension of the raw data and perform the similarity search of the sequential data. However, the computational complexity of this method is high, which will greatly reduce the efficiency of similarity search. The binary encoding generated by deep learning model improves the performance of time series data search. Attention-based model for time series similarity search using AIS Data segments can simultaneously acquire and learn the potential relationship between temporal information and parameter variables in time series data. Seq2Seq similarity search model based on attention mechanism combines the characteristics of time series data and Seq2Seq model, using the binary encoding generated by GRU encoder in Seq2Seq model to search for similarity on AIS data. The corresponding experimental results also show that this model based on the Attention mechanism has excellent similarity search performance on AIS data.
Based on the analysis of the development process of warship information system infrastructure for three generations, this paper summarizes its evolution rule, prospects the development trend of warship information system infrastructure from the perspectives of demand traction and technology promotion. It points out that the standardization is still the inner dynamics of the development of the information system infrastructure, and the platform will become the focus. It makes further analysis of the possible problems of the existing equipment, and proposes the ideas to solving problems.
Aiming at the problem of how aircraft can make rapid and intelligent decision according to real-time situation in air combat, this paper proposes an air combat behavior decision-making framework based on improved DDPG (ACBDF_DDPG). The main improvements in the framework are as follows: 1. An embedded artificial experience reward mechanism for dynamic targets is designed to alleviate the convergence difficulty caused by the huge state space and sparse reward in the training process of deep reinforcement learning algorithm; 2. Improve the actor network update mechanism in the framework to solve the problem of unstable model training caused by updating actor network when the evaluation effect of critical network is poor; 3. The priority sampling mechanism is adopted to ensure that the experience samples with high training value are fully utilized. Finally, a simulation experiment environment is built based on the MaCA, and the ablation experiment verifies the effectiveness and superiority of the improved mechanism proposed in this paper.
The offensive and defensive confrontation between the submarine and the surface ship formation is the key content of submarine combat research. How to ensure that the submarine survives and breaks through the joint blockade of the ship formation and anti-submarine helicopters is a test of the submarine command decision. To this end, in view of the asymmetry of the submarine-ship-helicopter confrontation scenario, the submarine agent is constructed from two aspects of deep reinforcement learning and rule inference, and two Proximal Policy Optimization (PPO) algorithm improvement mechanisms are proposed. It carries out mutual game confrontation and distributed training, and finally realizes the intelligent decision-making of submarines in the confrontation process. Related technical routes and algorithms have been implemented and verified on the wargaming platform. The improved algorithm has greatly improved the convergence speed and stability. The research on submarine intelligent command decision-making provides technical reference.
To improve the combat effectiveness of the ship in the support of land fire, a method of shipping route planning to travel to a stationary target on the shore is proposed. In this method, the projectile exterior ballistic equation and hit probability model of the ship’s naval gun are established, when the target information is obtained, it calculates the hit probability of the naval gun against the target at different target distances and side angles. Then the ship’s route is formed after connecting the ship’s position corresponding to the maximum hit probability at each distance. The results show on this route, the naval gun can maintain the maximum fire output against the enemy target at the current distance. When being discovered, our ship can quickly achieve damage to the mission target and evacuate, effectively enhance our ship’s combat capability and survivability.
In a real environment, it is difficult to realize the hit assessment of the naval gun system against high-speed maneuvering targets in the air. In this paper, a semi-physical simulation test system is constructed to simulate the battlefield environment and threat target route, drive the naval gun system to simulate strikes, collect the strike data in real time to calculate the projectile’s flight trajectory, and perform hit discrimination by fusion with theoretical target data. It analyses the target strike effect of the naval gun system in different typical target environments, different shooting sections, and different hit positions, which can provide an effective way for naval gun system combat capability evaluation and improvement design.
This paper studies a series of problems such as‘what data to collect’, ‘how to collect data’and‘how to analyze and process data’encountered by grass-roots units in carrying out data collection of equipment in-service assessment,and systematically expounds the classification,method and process of data collection of equipment in-service assessment. This paper gives a reasonable answer to the problems encountered in the work of in-service assessment data collection, studies the methods of quantitative index data processing, qualitative index data processing and abnormal index data processing, and puts forward a more reasonable data analysis and processing method, which provides a theoretical basis for the grass-roots units to carry out in-service assessment data collection in the next step.
Aiming at weapon equipment operational test index weighting problem, a new method based on AHP-EW has been proposed, on the basis of AHP, entropy weighting method is used to revise weight, which is based on group decision making theory, integrating subjectivity and objectivity, effectively solves the problem of low reliability of the results caused by the strong subjectivity of the traditional index weighting methods, gives full play to the collective wisdom, reflects the collective will, and provides a reference for the improvement of combat test evaluation.
Aiming at the characteristics that the naval battlefield information system capability evaluation tests involve many scenarios, equipments and subjects, and they are difficult to organize, a naval battlefield information system capability evaluation test design method based on mission profile is proposed, which divides the test requirements and test contents of naval battlefield information systems capability evaluation based on mission profiles. It is convenient to organize field tests, laboratory tests, simulation verifications, etc., standardize and clarify the tests organization processes, and maximize the reuse of conclusions.
With the in-depth development of the informationization construction of army equipment in the new era, the demand for equipment support of new army equipment of our army has been significantly increased, and higher requirements have been put forward for the construction and development of army equipment support business. In order to effectively improve the army’s equipment support capability and optimize the development path of equipment support business, SWOT is firstly used to analyze the constraints affecting the development of the army’s equipment support business at the present stage, and a SWOT analysis matrix is established based on the four aspects of "strengths, weaknesses, opportunities and threats".Secondly, combining with the quantitative AHP model, the factors are analyzed, quantified and ranked, and then the SWOT strategic strength coordinate system is constructed to determine the WO aggressive strategy, and the development countermeasures in policy, capital, talent and technology are proposed to provide reference for the development planning of equipment support business under the new system.
