A combat situation analysis and target system action prediction method and system based on a large language model
By combining manual and automatic analysis based on a large language model, and integrating simulation models and the experience of commanders, the problem of insufficient training data for neural networks was solved, enabling accurate adjustment of enemy target grouping and action prediction, and improving the accuracy of combat situation analysis.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- BEIJING GUANTIAN TECH CO LTD
- Filing Date
- 2026-02-11
- Publication Date
- 2026-05-29
AI Technical Summary
In existing technologies, when using neural networks for combat situation analysis, the limited availability of training data due to confidentiality and authenticity issues makes it impossible to effectively train the model, resulting in inaccurate analysis results.
The method employs a combination of manual and automatic analysis based on a large language model. By acquiring a set of enemy targets, performing cluster analysis and adjustments, using simulation models for deduction, and combining the experience of commanders with language analysis instructions for correction, it achieves the adjustment of enemy target groupings and the prediction of strike destinations.
With limited data, relatively accurate predictions of enemy target actions can be obtained, improving the accuracy of combat situation analysis and decision support capabilities.
Smart Images
Figure CN122114765A_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of situation analysis technology, and in particular to a method and system for operational situation analysis and target system action prediction based on a large language model. Background Technology
[0002] Operational situation analysis and target system action prediction are core prerequisites for military command and decision-making. They are key technological links connecting battlefield intelligence perception and command and decision-making. The purpose is to achieve accurate understanding of the current battlefield status and prediction of the future actions of the enemy target system by integrating, analyzing and extrapolating multi-source battlefield information, ultimately providing a scientific basis for operational decision-making.
[0003] Currently, we are trying to use neural networks for analysis. The advantage of neural network analysis is that it can mine hidden correlations from massive heterogeneous data, achieve adaptive learning and accurate prediction, and has significant performance and efficiency improvements compared to traditional statistical models and rule engines.
[0004] However, a practical problem to consider is that most of the real data used to train this type of neural network is currently kept confidential, resulting in a very limited amount of data that can be used for training. Furthermore, the lack of authenticity of the data makes it impossible to effectively train the model. Summary of the Invention
[0005] This application provides a method and system for combat situation analysis and target system action prediction based on a large language model. It combines manual and automatic analysis to perform combat situation analysis and target system action prediction, which can obtain relatively accurate results based on limited data.
[0006] The above-mentioned objective of this application is achieved through the following technical solution: Firstly, this application provides a method for operational situation analysis and target system action prediction based on a large language model, including: Obtain the set of enemy targets within the coverage area; the set of enemy targets includes multiple enemy targets. Cluster analysis is performed on enemy targets to obtain enemy target groups; The received language analysis instructions are parsed to obtain the adjustment results; The enemy target groups were adjusted based on the results of the adjustments; The target destinations of enemy target groups are simulated to obtain action prediction results.
[0007] In one possible implementation of the first aspect, cluster analysis of enemy targets includes: Obtain our combat groups and the combat capability parameters of each combat group; The enemy targets are divided according to the combat capability parameters and set objectives of each combat group, resulting in a first grouping of the enemy targets; Cluster analysis is performed on the initial grouping of enemy targets to obtain enemy target groups.
[0008] In one possible implementation of the first aspect, cluster analysis of a single grouping of enemy targets includes: A dynamic mesh is created using the coordinates of the enemy target, and each grid in the dynamic mesh has the same number of edges. Identify the local independent regions of the dynamic mesh, with at least one local independent region. Determine the dynamic grid region associated with the local independent region; Determine the overlap between the dynamic grid region and a single group of enemy targets; When the overlap between the dynamic grid region and a group of enemy targets exceeds a set overlap, the enemy targets corresponding to the dynamic grid region are grouped as enemy targets.
[0009] In one possible implementation of the first aspect, determining the local independent regions of the dynamic mesh includes: Determine the area value of each grid cell in the dynamic grid; The grids are grouped according to their area values to obtain grid groups, and the grids in each group have the same or similar areas. The dynamic grid is filtered by grid grouping to obtain a dynamic grid cluster. The number of grids included in the dynamic grid cluster is greater than or equal to the set number. Dynamic mesh clusters are treated as local, independent regions of the dynamic mesh.
[0010] In one possible implementation of the first aspect, determining the area value of each grid in the dynamic grid also includes correcting the area value of each grid using a topographic map.
[0011] In one possible implementation of the first aspect, after obtaining the local independent regions of the dynamic mesh, the method further includes verifying the enemy target groups. Verifying the enemy target groups includes: Track dynamic grid aggregation over time; When the continuous occurrence time or cumulative occurrence time of dynamic grid clusters is greater than or equal to a set time, the enemy target group corresponding to the local independent area passes the verification.
[0012] In one possible implementation of the first aspect, when tracking the dynamic mesh cluster over time, it is allowed that the number of enemy targets included in the dynamic mesh cluster changes, and the number of enemy targets that change is less than or equal to the allowed number.
[0013] Secondly, this application provides a combat situation analysis and target system action prediction device based on a large language model, comprising: The data acquisition unit is used to acquire a set of enemy targets within the coverage area, which includes multiple enemy targets; The first analysis unit is used to perform cluster analysis on enemy targets to obtain enemy target groups; The second analysis unit is used to parse the received language analysis instructions and obtain the adjustment results; The adjustment unit is used to adjust the grouping of enemy targets based on the adjustment results; The action prediction unit is used to simulate the target destinations of enemy target groups and obtain action prediction results.
[0014] Thirdly, this application provides a combat situation analysis and target system action prediction system based on a large language model, the system comprising: One or more memories for storing instructions; and One or more processors are configured to call and execute the instructions from the memory to perform the methods described in the first aspect and any possible implementation thereof.
[0015] Fourthly, this application provides a computer-readable storage medium, the computer-readable storage medium comprising: The program, when run by a processor, is executed as described in the first aspect and any possible implementation thereof.
[0016] Fifthly, this application provides a computer program product, including program instructions that, when run by a computing device, execute the method described in the first aspect and any possible implementation thereof.
[0017] Sixthly, this application provides a chip system including a processor for implementing the functions involved in the foregoing aspects, such as generating, receiving, transmitting, or processing the data and / or information involved in the foregoing methods.
[0018] This chip system can consist of chips or include chips and other discrete components.
[0019] In one possible design, the chip system also includes a memory for storing necessary program instructions and data. The processor and the memory can be decoupled and located on different devices, connected via wired or wireless means, or the processor and the memory can be coupled to the same device. Attached Figure Description
[0020] Figure 1 This is a flowchart illustrating the steps of a combat situation analysis and target system action prediction method based on a large language model provided in this application.
[0021] Figure 2 This is a schematic diagram provided in this application of a commander adjusting the grouping of enemy targets.
[0022] Figure 3 This is a schematic diagram illustrating the division of enemy targets as provided in this application.
[0023] Figure 4 This is a schematic diagram of a dynamic mesh created using the coordinates of an enemy target, as provided in this application.
[0024] Figure 5 This is a schematic diagram of a dynamic mesh aggregation method provided in this application. Detailed Implementation
[0025] The technical solutions in this application will be further described in detail below with reference to the accompanying drawings.
[0026] This application discloses a method for operational situation analysis and target system action prediction based on a large language model. Please refer to [link / reference]. Figure 1 In some examples, the operational situation analysis and target system action prediction method based on large language models disclosed in this application includes the following steps: S101, Obtain the set of enemy targets within the coverage area, the set of enemy targets includes multiple enemy targets; S102, perform cluster analysis on enemy targets to obtain enemy target groups; S103, parse the received language analysis instructions to obtain the adjustment results; S104, Adjust the enemy target grouping based on the adjustment results; S105 simulates the target destinations of enemy target groups to obtain action prediction results.
[0027] First, let's introduce the relevant battlefield environment.
[0028] The introduction of drones into the battlefield has brought about a qualitative change in the form of warfare. Specifically, compared with traditional warfare, drones, with their core advantages of "zero casualties, high efficiency, low cost, and wide coverage," have achieved a disruptive reconstruction of the form of warfare in all dimensions, from tactics to strategy, from organization to ethics, and from offense and defense to technology, thus accelerating the evolution of warfare towards unmanned, intelligent, and asymmetric warfare.
[0029] In traditional warfare, coordination from target discovery to fire strike takes hours, and pilots face a high risk of injury or death. Drones compress the kill chain to minutes, achieving "detect and destroy," and remote control completely avoids personnel casualties, changing the political and public opinion cost logic of warfare and significantly lowering the decision-making threshold. Low-cost loitering munitions / FPV drones create cost asymmetry, consuming high value with low value, subverting the traditional equipment cost-exchange ratio.
[0030] In traditional warfare, the front and rear are clearly defined, with the rear being relatively safe. However, the all-encompassing coverage of drones exposes deep targets (command centers, logistical nodes) to continuous attack risks, making the rear no longer safe. Low-altitude / micro-platform drones can penetrate traditional air defense blind spots, and continuous day and night monitoring drastically reduces the effectiveness of camouflage and night raids, resulting in unprecedented battlefield transparency.
[0031] Taking the current war as an example, the addition of drones has made large-scale troop concentrations extremely dangerous. Troops must be in a state of concealment or rapid movement. From the perspective of positional warfare, the attack method has begun to shift to a flexible attack method in which small groups of troops move in a dispersed manner and then combine to act after reaching their destination.
[0032] The technical solution in this application is designed to address scenarios where small groups of troops move in a dispersed manner. Specifically, in step S101, the set of enemy targets within the coverage area is first obtained. The set of enemy targets includes multiple enemy targets. Then, in step S102, the enemy targets are clustered to obtain enemy target groups. At this time, an enemy target group includes one enemy target or multiple enemy targets.
[0033] Here, coverage refers to the coverage area using one's own reconnaissance methods, which include high / low-altitude drone reconnaissance, satellite reconnaissance, and personnel reconnaissance, etc., and are not restricted here.
[0034] In step S103, the received language analysis instructions are parsed to obtain the adjustment results. Here, correlation analysis is performed based on the instructions issued by the commander. Next, in step S104, the enemy target groups are adjusted according to the adjustment results. This adjustment refers to determining which specific enemy target group each target is assigned to.
[0035] The language analysis instructions here are given by the commander. The commander adjusts the enemy target grouping obtained in step S102 based on the intelligence collected and his own capabilities. The reason for the adjustment is that the method in step S102 is a predetermined analysis method. Its process can be regarded as a derivation process based on probability statistics. This process requires a large amount of data for learning. In the case of missing data in the early stage, the accuracy of this processing is limited.
[0036] For example, enemy targets can be numbered, and the commanding officer could instruct them to associate number 2 with number 5, such as... Figure 2 As shown, Figure 2 The dots in the diagram represent enemy targets, and the areas within the dashed boxes represent groups of enemy targets belonging to the same group.
[0037] As mentioned in the background section of this application, the data currently available for analysis is very limited, resulting in a probability of deviation in the analysis results. Therefore, it is necessary for the command personnel to make corrections. The purpose of the corrections is twofold: one is to quickly determine the groupings, and the other is to conduct experiments by intentionally changing the groupings to probe the true intentions of the enemy targets.
[0038] For example, after making corrections, if the commander discovers an enemy target group, they can deduce the enemy target group's arrival location and attempt to strike our targets based on the intelligence gathered and the enemy target group's movement path and firepower configuration. This part is part of the commander's job, and the description here is only for clarification.
[0039] Finally, in step S105, the strike destinations of the enemy target groups are simulated to obtain the action prediction results. The simulation method here is based on strike capabilities, and the specific process is as follows: Based on the collected intelligence, the number of targets (including the number of personnel, the number of equipment, and the type of equipment) and the target parameters (the equipment's striking capability, the probability of hitting, the speed of attack, etc.) are determined, and then these data are input into the simulation model for simulation.
[0040] In some possible implementations, the simulation model uses: Multi-Agent System (MAS) models drones, artillery, air defense systems, etc., as autonomous agents. Each agent has perception, decision-making, and action capabilities, and simulates cooperation and confrontation through rule engines / reinforcement learning.
[0041] The wargaming simulation system (including a rule engine and digital twin) uses a rule base (such as D2T2E kill chain rules) + GIS + physics engine to support turn-based / real-time simulations and visualize battlefield situation and damage effects.
[0042] In some cases, the specific methods for performing cluster analysis on enemy targets are as follows: S201, Obtain our combat groups and the combat capability parameters of each combat group; S202, based on the combat capability parameters and set objectives of each combat group, enemy targets are divided into groups, resulting in a first grouping of enemy targets; S203, perform cluster analysis on the initial grouping of enemy targets to obtain enemy target groups.
[0043] Please see Figure 3 Steps S201 to S203 involve classifying enemy targets based on the combat capability parameters and set objectives of each combat group, with reference to... Figure 3 The dotted line in the diagram represents the combat capability parameter, which refers to the combat capability of a friendly combat group. For example, the combat capability of a combatant carrying an X1 weapon is set to 1, the combat capability of a combatant carrying an X2 weapon is set to 2, and so on. Here, "combatant" refers to any person using a weapon. If long-range strikes are possible, then long-range strikes are also quantified.
[0044] The classification of enemy targets and the number of enemy targets in a group are generally given by the commander (e.g., via voice) based on experience. This is mainly because the influencing parameters differ under different battlefield situations, making it impossible to form a unified judgment standard. The experience of the commander is required here.
[0045] Setting targets refers to the intentions of enemy target groups towards our combat groups. These intentions include annihilation, containment, harassment, and interception. Here, there are two ways to classify enemy targets based on the combat capability parameters of each combat group and the set targets. One is to use the simulation model described above to perform a deduction. This will yield a probability value. Taking annihilation as an example, an annihilation probability value will be obtained. When this annihilation probability value is greater than a reference value (e.g., 80%), it means that the current grouping of enemy targets is reasonable. Otherwise, it is necessary to continue to classify enemy targets based on the combat capability parameters of each combat group and the set targets until the conditions are met.
[0046] Another approach is to directly compare the values and add a safety value. For example, if the combat capability parameter of our combat group is 1, and the safety value is set to 1.5, then in order to meet the annihilation objective, the combat capability parameter of the enemy target group should be greater than or equal to 1.5.
[0047] Of course, factors such as time also need to be considered at this point. The safety value can be adjusted according to specific requirements and can be adjusted by the commander via voice. For example, if the commander believes that it will take half an hour to strike one of our combat groups and requires at least three times the combat capability parameters, then the safety value can be set to 3 or a value greater than 3.
[0048] The division of enemy targets based on the combat capability parameters and set objectives of each combat group is based on the combat capability and set objectives of a friendly combat group. This division results in a grouping of enemy targets.
[0049] However, cluster analysis is required for the initial grouping of enemy targets. The specific steps for cluster analysis are as follows: A dynamic mesh is created using the coordinates of the enemy target, and each grid in the dynamic mesh has the same number of edges. Identify the local independent regions of the dynamic mesh, with at least one local independent region. Determine the dynamic grid region associated with the local independent region; Determine the overlap between the dynamic grid region and a single group of enemy targets; When the overlap between the dynamic grid region and a group of enemy targets exceeds a set overlap, the enemy targets corresponding to the dynamic grid region are grouped as enemy targets.
[0050] Specifically, a dynamic grid is first established using the coordinates of the enemy target, such as... Figure 4 As shown, each grid in the dynamic mesh has the same number of edges, typically three. Next, the local independent regions of the dynamic mesh are determined. There can be one or more local independent regions.
[0051] Next, the dynamic grid region associated with the local independent region is determined, and then the overlap between the dynamic grid region and the first group of enemy targets is determined. Specifically, the basis of this process is that when multiple enemy targets attack a friendly combat group, there will be certain correlations in their movement and distance. By overlapping the dynamic grid region with the first group of enemy targets, the inherent correlations between multiple enemy targets can be discovered.
[0052] The specific method for determining the local independent regions of a dynamic mesh is as follows: Determine the area value of each grid cell in the dynamic grid; The grids are grouped according to their area values to obtain grid groups, and the grids in each group have the same or similar areas. The dynamic grid is filtered by grid grouping to obtain a dynamic grid cluster. The number of grids included in the dynamic grid cluster is greater than or equal to the set number. Dynamic mesh clusters are treated as local, independent regions of the dynamic mesh.
[0053] The above method first places grids with the same or similar areas into the same grid group. Then, it searches the dynamic grid to see if there is a certain degree of clustering among the grids in the same group. For example, if the minimum value is set to 3, then three grids in the same group need to be adjacent. Figure 5 As shown.
[0054] The method for grouping grids based on their area values is as follows: First, sort the area values from smallest to largest to obtain an area value sequence. The sorting position of the area values is the horizontal axis, and the size of the area values is the vertical axis. This yields discrete points in the coordinate system, which are then processed using the DBSCAN (density clustering) method.
[0055] Finally, the dynamic grid is aggregated as a local independent region of the dynamic grid. When determining the area value of each grid in the dynamic grid, the area value of each grid is also corrected using a terrain map. Specifically, the terrain directly affects the distance between two enemy targets. The reconnaissance methods described above are generally based on planar display and ignore the influence of terrain. Specifically, the terrain directly affects the assembly and infiltration of tactical units, as well as fire support (the fire coverage of the same terrain has a high overlap rate, and the support is more accurate) and communication (when there is no terrain obstruction, radio and radar signal transmission is stable, and the situational awareness of the two units is highly synchronized).
[0056] Therefore, it is necessary to use a topographic map to correct the area value of each grid. The correction method is to calculate the actual distance between two enemy targets based on the terrain, or to directly use a coefficient for correction. For example, a coefficient can be directly assigned to a certain terrain. The actual distance between two enemy targets = the map distance between the two enemy targets * the terrain coefficient.
[0057] In some cases, after obtaining the density center of the dynamic grid, the process also includes verifying the enemy target groups, as follows: Track dynamic grid aggregation over time; When the continuous occurrence time or cumulative occurrence time of dynamic grid clusters is greater than or equal to a set time, the enemy target group corresponding to the density center passes the verification.
[0058] The above method is based on the continuous or cumulative occurrence time of dynamic grid clusters. The set time here is a manually input reference value. When the continuous or cumulative occurrence time of dynamic grid clusters is greater than or equal to the set time, the enemy target group corresponding to the density center passes the verification.
[0059] In some possible implementations, when tracking dynamic mesh clusters over time, the number of enemy targets included in the dynamic mesh clusters is allowed to change, with the number of enemy targets that change being less than or equal to the allowed number.
[0060] This approach allows for the possibility of inaccurate judgments in the early stages due to limited basic conditions. Its core principle is that as long as the changes in the dynamic grid cluster are within the allowable range, the dynamic grid cluster will be continuously tracked. In some possible implementations, the number of enemy targets that have changed is less than or equal to the allowable number, which is generally 10%-15% of the total number of enemy targets in the dynamic grid cluster.
[0061] This application also provides a combat situation analysis and target system action prediction device based on a large language model, including: The data acquisition unit is used to acquire a set of enemy targets within the coverage area, which includes multiple enemy targets; The first analysis unit is used to perform cluster analysis on enemy targets to obtain enemy target groups; The second analysis unit is used to parse the received language analysis instructions and obtain the adjustment results; The adjustment unit is used to adjust the grouping of enemy targets based on the adjustment results; The action prediction unit is used to simulate the target destinations of enemy target groups and obtain action prediction results.
[0062] Furthermore, cluster analysis of enemy targets includes: Obtain our combat groups and the combat capability parameters of each combat group; The enemy targets are divided according to the combat capability parameters and set objectives of each combat group, resulting in a first grouping of the enemy targets; Cluster analysis is performed on the initial grouping of enemy targets to obtain enemy target groups.
[0063] Furthermore, cluster analysis of a single group of enemy targets includes: A dynamic mesh is created using the coordinates of the enemy target, and each grid in the dynamic mesh has the same number of edges. Identify the local independent regions of the dynamic mesh, with at least one local independent region. Determine the dynamic grid region associated with the local independent region; Determine the overlap between the dynamic grid region and a single group of enemy targets; When the overlap between the dynamic grid region and a group of enemy targets exceeds a set overlap, the enemy targets corresponding to the dynamic grid region are grouped as enemy targets.
[0064] Furthermore, determining the local independent regions of the dynamic mesh includes: Determine the area value of each grid cell in the dynamic grid; The grids are grouped according to their area values to obtain grid groups, and the grids in each group have the same or similar areas. The dynamic grid is filtered by grid grouping to obtain a dynamic grid cluster. The number of grids included in the dynamic grid cluster is greater than or equal to the set number. Dynamic mesh clusters are treated as local, independent regions of the dynamic mesh.
[0065] Furthermore, determining the area value of each grid in the dynamic grid also includes correcting the area value of each grid using a topographic map.
[0066] Furthermore, after obtaining the local independent regions of the dynamic grid, the process also includes verifying enemy target groups. Verification of enemy target groups includes: Track dynamic grid aggregation over time; When the continuous occurrence time or cumulative occurrence time of dynamic grid clusters is greater than or equal to a set time, the enemy target group corresponding to the local independent area passes the verification.
[0067] Furthermore, when tracking dynamic grid clusters over time, the number of enemy targets included in the dynamic grid clusters is allowed to change, with the number of enemy targets that change being less than or equal to the allowed number.
[0068] A combat situation analysis and target system action prediction device based on a large language model, comprising: The data acquisition unit is used to acquire a set of enemy targets within the coverage area, which includes multiple enemy targets; The first analysis unit is used to perform cluster analysis on enemy targets to obtain enemy target groups; The second analysis unit is used to parse the received language analysis instructions and obtain the adjustment results; The adjustment unit is used to adjust the grouping of enemy targets based on the adjustment results; The action prediction unit is used to simulate the target destinations of enemy target groups and obtain action prediction results.
[0069] A combat situation analysis and target system action prediction system based on a large language model, characterized in that the system comprises: One or more memories for storing instructions; and One or more processors are configured to retrieve and execute the instructions from the memory to perform the method as described in any one of claims 1 to 7.
[0070] A computer-readable storage medium, characterized in that the computer-readable storage medium comprises: The program, when run by the processor, executes the method as described in any one of claims 1 to 7.
[0071] In one example, the unit in any of the above devices may be one or more integrated circuits configured to implement the above methods, such as one or more application-specific integrated circuits (ASICs), or one or more digital signal processors (DSPs), or one or more field-programmable gate arrays (FPGAs), or a combination of at least two of these integrated circuit forms.
[0072] For example, when the units in the device can be implemented through a processing element scheduler, the processing element can be a general-purpose processor, such as a central processing unit (CPU) or other processor capable of calling programs. Alternatively, these units can be integrated together to form a system-on-a-chip (SOC).
[0073] In this application, various objects such as messages / information / devices / network elements / systems / apparatus / actions / operations / processes / concepts may be named. It is understood that these specific names do not constitute a limitation on the relevant objects. The names may be changed depending on the scenario, context, or usage habits. The understanding of the technical meaning of the technical terms in this application should be mainly determined from their functions and technical effects embodied / performed in the technical solution.
[0074] Those skilled in the art will understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.
[0075] In the several embodiments provided in this application, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between apparatuses or units may be electrical, mechanical, or other forms.
[0076] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.
[0077] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.
[0078] It should also be understood that in the various embodiments of this application, the terms "first," "second," etc., are merely to indicate that multiple objects are different. For example, a first time window and a second time window are only to indicate different time windows. They should not have any effect on the time windows themselves, and the aforementioned terms "first," "second," etc., should not impose any limitations on the embodiments of this application.
[0079] It should also be understood that, in the various embodiments of this application, unless otherwise specified or in case of logical conflict, the terms and / or descriptions between different embodiments are consistent and can be referenced by each other, and the technical features in different embodiments can be combined to form new embodiments according to their inherent logical relationships.
[0080] If the aforementioned functions are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a computer-readable storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned computer-readable storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.
[0081] This application also provides a combat situation analysis and target system action prediction system based on a large language model, the system comprising: One or more memories for storing instructions; and One or more processors are configured to retrieve and execute the instructions from the memory, performing the methods described above.
[0082] This application also provides a computer program product including instructions that, when executed, cause the terminal device and the network device to perform operations corresponding to the methods described above.
[0083] This application also provides a chip system including a processor for implementing the functions involved in the above description, such as generating, receiving, transmitting, or processing the data and / or information involved in the above methods.
[0084] This chip system can consist of chips or include chips and other discrete components.
[0085] The processor mentioned above can be a CPU, a microprocessor, an ASIC, or one or more integrated circuits that execute a program to control the method of transmitting the feedback information described above.
[0086] In one possible design, the chip system also includes a memory for storing necessary program instructions and data. The processor and the memory can be decoupled and located on different devices, connected via wired or wireless means to support the chip system in implementing the various functions described in the above embodiments. Alternatively, the processor and the memory can also be coupled to the same device.
[0087] Optionally, the computer instructions are stored in memory.
[0088] Optionally, the memory can be a storage unit within the chip, such as a register or cache. Alternatively, the memory can be a storage unit located outside the chip within the terminal, such as a ROM or other types of static storage devices that can store static information and instructions, such as RAM.
[0089] It is understood that the memory in this application may be volatile memory or non-volatile memory, or may include both volatile and non-volatile memory.
[0090] Non-volatile memory can be ROM, programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), or flash memory.
[0091] Volatile memory can be RAM, which is used as an external cache. There are many different types of RAM, such as static random access memory (SRAM), dynamic random access memory (DRAM), synchronous dynamic random access memory (SDRAM), double data rate synchronous dynamic random access memory (DDR SDRAM), enhanced synchronous dynamic random access memory (ESDRAM), synchronous link dynamic random access memory (SLDRAM), and direct memory bus random access memory.
[0092] The embodiments described in this specific implementation are preferred embodiments of this application and are not intended to limit the scope of protection of this application. Therefore, all equivalent changes made in accordance with the structure, shape and principle of this application should be covered within the scope of protection of this application.
Claims
1. A method for combat situation analysis and target system action prediction based on a large language model, characterized in that, include: Obtain the set of enemy targets within the coverage area; the set of enemy targets includes multiple enemy targets. Cluster analysis is performed on enemy targets to obtain enemy target groups; The received language analysis instructions are parsed to obtain the adjustment results; The enemy target groups were adjusted based on the results of the adjustments; The target destinations of enemy target groups are simulated to obtain action prediction results.
2. The method for combat situation analysis and target system action prediction based on a large language model according to claim 1, characterized in that, Cluster analysis of enemy targets includes: Obtain our combat groups and the combat capability parameters of each combat group; The enemy targets are divided according to the combat capability parameters and set objectives of each combat group, resulting in a first grouping of the enemy targets; Cluster analysis is performed on the initial grouping of enemy targets to obtain enemy target groups.
3. The method for combat situation analysis and target system action prediction based on a large language model according to claim 2, characterized in that, Cluster analysis of a single group of enemy targets includes: A dynamic mesh is created using the coordinates of the enemy target, and each grid in the dynamic mesh has the same number of edges. Identify the local independent regions of the dynamic mesh, with at least one local independent region. Determine the dynamic grid region associated with the local independent region; Determine the overlap between the dynamic grid region and a single group of enemy targets; When the overlap between the dynamic grid region and a group of enemy targets exceeds a set overlap, the enemy targets corresponding to the dynamic grid region are grouped as enemy targets.
4. The method for combat situation analysis and target system action prediction based on a large language model according to claim 3, characterized in that, Determining the local independent regions of a dynamic mesh includes: Determine the area value of each grid cell in the dynamic grid; The grids are grouped according to their area values to obtain grid groups, and the grids in each group have the same or similar areas. The dynamic grid is filtered by grid grouping to obtain a dynamic grid cluster. The number of grids included in the dynamic grid cluster is greater than or equal to the set number. Dynamic mesh clusters are treated as local, independent regions of the dynamic mesh.
5. The method for combat situation analysis and target system action prediction based on a large language model according to claim 4, characterized in that, Determining the area value of each grid in the dynamic grid also includes correcting the area value of each grid using a topographic map.
6. The method for combat situation analysis and target system action prediction based on a large language model according to claim 4, characterized in that, After obtaining the local independent regions of the dynamic mesh, the process also includes verifying enemy target groups. Verification of enemy target groups includes: Track dynamic grid aggregation over time; When the continuous occurrence time or cumulative occurrence time of dynamic grid clusters is greater than or equal to a set time, the enemy target group corresponding to the local independent area passes the verification.
7. The method for combat situation analysis and target system action prediction based on a large language model according to claim 6, characterized in that, When tracking dynamic mesh clusters over time, changes in the number of enemy targets included in the dynamic mesh cluster are allowed, provided that the number of enemy targets that have changed is less than or equal to the allowed number.
8. A combat situation analysis and target system action prediction device based on a large language model, characterized in that, include: The data acquisition unit is used to acquire a set of enemy targets within the coverage area, which includes multiple enemy targets; The first analysis unit is used to perform cluster analysis on enemy targets to obtain enemy target groups; The second analysis unit is used to parse the received language analysis instructions and obtain the adjustment results; The adjustment unit is used to adjust the grouping of enemy targets based on the adjustment results; The action prediction unit is used to simulate the target destinations of enemy target groups and obtain action prediction results.
9. A combat situation analysis and target system action prediction system based on a large language model, characterized in that, The system includes: One or more memories for storing instructions; and One or more processors are configured to retrieve and execute the instructions from the memory to perform the method as described in any one of claims 1 to 7.
10. A computer-readable storage medium, characterized in that, The computer-readable storage medium includes: The program, when run by the processor, executes the method as described in any one of claims 1 to 7.