Multi-step confrontation simulation model construction method, system and equipment

Abstract

The invention discloses a multi-step confrontation simulation model construction method, which comprises the following steps of S1, selecting a plurality of sub-models from an existing model set according to a confrontation task, and distributing an initial simulation step for each sub-model; s2, acquiring running states of the sub-models in a simulation process, and dynamically adjusting the initial simulation step length of each sub-model according to the running states to obtain a final simulation step length; and S3, establishing a time synchronization and simulation data interpolation mechanism by taking the minimum final simulation step size in all the sub-models as a reference step size according to the simulation advancing sequence and the causal relationship of all the sub-models, and obtaining an adversarial simulation model. According to the method, the problems of resource control and scheduling loss in fixed step size synchronous simulation propulsion and the problems of data output imbalance and precision error accumulation caused by asynchronous simulation propulsion of fixed mixed step size scheduling in the prior art are solved, and the operation efficiency and the simulation precision are improved.

Description

Technical Field

[0001] This invention relates to the field of adversarial simulation technology, and in particular to a method, system and device for constructing a multi-step adversarial simulation model. Background Technology

[0002] With the increasing sophistication of modern warfare, system-on-systems simulation is playing an increasingly important role in the military field. Traditional simulation methods typically employ a fixed step size, using the same step size across the entire system. This leads to a trade-off between computational efficiency and simulation accuracy. Systems with rapidly changing loads require smaller simulation step sizes to achieve convergence, while systems with slower load changes require larger step sizes. Choosing a smaller step size can improve simulation accuracy but significantly reduces computational efficiency; conversely, choosing a larger step size can improve computational efficiency but results in larger numerical errors in the simulation results, reducing accuracy.

[0003] Furthermore, in large-scale system-on-system simulations, multiple subsystems are typically involved, each of which may have different time-scale characteristics. Traditional fixed-step simulation methods struggle to accurately reflect the spatiotemporal consistency between the subsystems, thus affecting the reliability of the simulation results.

[0004] Currently, there are two main solutions in the industry to address the trade-off between computational efficiency and simulation accuracy: One is Chinese invention patent CN118013734A, which discloses a multi-step real-time simulation method for integrated energy systems. This method reduces simulation computation time while reflecting the real operating conditions of integrated energy systems. However, when decomposing the dynamic model of the integrated energy system, this method does not consider the characteristics and needs of different energy systems and cannot perform more flexible subsystem partitioning. The other is Chinese invention patent CN117973093A, which discloses a simulation step size determination method, including determining the minimum main step size and the minimum secondary step size. This method can efficiently determine the final simulation step size while saving computational resources. However, when determining the minimum main step size, it does not consider introducing more iteration conditions, making it difficult to improve the accuracy and stability of the determined minimum main step size.

[0005] In summary, the existing technology has the following drawbacks: 1. Traditional synchronous simulation with fixed step size suffers from resource management and scheduling overhead, resulting in low computational efficiency; while asynchronous simulation with fixed hybrid step size scheduling leads to unbalanced data output and accumulation of accuracy errors, resulting in low simulation accuracy. 2. Data frequency imbalance in high-performance simulation models: When the simulation resolution is too low, the data output frequency will be too low. When the data is random or has a large range of variation, the application of preprocessing methods such as data smoothing will bring about a large prediction deviation. On the other hand, when the simulation resolution is too high, it will increase the memory processing burden and affect the overall computing performance of parallel simulation. Summary of the Invention

[0006] This invention discloses a method for constructing a multi-step adversarial simulation model, which mainly solves the technical problems in the prior art, such as the inability to balance the computational efficiency and simulation accuracy of synchronous simulation under a fixed step size, and the difficulty in ensuring the spatiotemporal consistency of simulation models with different step sizes.

[0007] To address the aforementioned technical problems, this invention discloses a method for constructing a multi-step adversarial simulation model, the method comprising the following steps: S1. Select multiple sub-models from the existing model set according to the adversarial task, and assign an initial simulation step size to each sub-model; S2. Obtain the running status of the sub-model during the simulation process, and dynamically adjust the initial simulation step size of each sub-model according to the running status to obtain the final simulation step size; S3. Based on the simulation progression order and causal relationship of all the sub-models, the smallest final simulation step size among all the sub-models is used as the reference step size to establish a time synchronization and simulation data interpolation mechanism to obtain the adversarial simulation model.

[0008] Further, in step S1, multiple sub-models are selected from the existing model set according to the adversarial task, and an initial simulation step size is assigned to each sub-model, including: S11. Perform task analysis on the adversarial task to obtain multiple combat units, and select the sub-model corresponding to each combat unit from the existing model set; S12. Analyze and obtain the load change rate for each combat unit, and determine the time scale of each sub-model based on the load change rate; S13. Assign an initial simulation step size to each of the sub-models according to the time scale.

[0009] Furthermore, in step S13, an initial simulation step size is assigned to each of the sub-models according to the time scale, including: S131. Set a baseline simulation step size corresponding to the baseline time scale; S132. Calculate the multiple relationship between the reference time scale and the time scale of each sub-model; S133. The product of the calculated multiple relationship and the baseline simulation step size is used as the initial simulation step size for each sub-model.

[0010] Further, in step S2, obtaining the running state of the sub-model during the simulation includes: S21. Insert logs, callback functions, or state variable update code at the key nodes of each sub-model; S22. During the simulation, the state information output by each sub-model is acquired in real time, and the running state of each sub-model is obtained by analyzing the state information. The running state includes combat intensity score and error accumulation.

[0011] Furthermore, in step S2, the initial simulation step size of each sub-model is dynamically adjusted according to the running state to obtain the final simulation step size, including: S23. Calculate the combat intensity range of each combat unit based on the environmental factors of the confrontation mission and the combat capability, personnel status and resource support of each combat unit in the confrontation mission. S24. Set an error accumulation threshold for each of the sub-models; S25. During the simulation, obtain the combat intensity score and error accumulation of each sub-model; S26. When the cumulative error of the sub-model is less than the cumulative error threshold, the combat intensity score is used as a constraint within the combat intensity range. The initial simulation step size is gradually increased by any one of the variable look-ahead simulation advancement algorithm, distance accuracy simulation advancement algorithm, and adaptive algorithm to obtain the final simulation step size. S27. When the cumulative error of the sub-model is greater than or equal to the error accumulation threshold, the initial simulation step size is gradually reduced by any one of the variable look-ahead simulation advancement algorithm, distance accuracy simulation advancement algorithm, and adaptive algorithm, with the combat intensity score within the combat intensity range as a constraint, until the final simulation step size satisfies the condition that the cumulative error is less than the error accumulation threshold.

[0012] Furthermore, in steps S26 and S27, the variable look-ahead simulation advancement method is as follows: predict the look-ahead quantity of the next simulation after the current simulation of the sub-model by using empirical parameters or statistical methods, and dynamically adjust the current simulation step size corresponding to the current simulation based on the look-ahead quantity to obtain the simulation step size of the next simulation.

[0013] The distance accuracy simulation advancement method is as follows: obtain the error compensation distance accuracy value of the sub-model, and dynamically adjust the simulation step size of the next simulation after each simulation of the sub-model based on the error compensation distance accuracy value to obtain the simulation step size of the next simulation, wherein the error compensation distance accuracy value is the parameter resolution value of the sub-model in a single cycle.

[0014] Further, in step S3, based on the simulation progression order and causal relationship of all the sub-models, a time synchronization and simulation data interpolation mechanism is established using the smallest final simulation step size among all the sub-models as the reference step size, including: S31. Based on the analysis results of the adversarial task, establish the causal relationship between all the sub-models, and establish a dependency graph based on the causal relationship; S32. Based on the dependency graph, determine the simulation progression order of each sub-model according to the topology sorting algorithm; S33. According to the simulation progress order, the smallest final simulation step size among all the sub-models is used as the reference step size. The simulation time is divided into multiple synchronization points. For the sub-models whose final simulation step size is greater than the reference step size, data interpolation is performed at each synchronization point to establish a time synchronization and simulation data interpolation mechanism.

[0015] Furthermore, the method also includes: S4. A distributed parallel computing architecture is adopted, and computing nodes are allocated to each sub-model in the adversarial simulation model for parallel simulation computing.

[0016] The present invention also discloses a multi-step adversarial simulation model construction system, including a model partitioning module, a simulation step size allocation module, a simulation step size correction module, and a time synchronization and simulation data interpolation mechanism establishment module.

[0017] The model partitioning module is used to select multiple sub-models from the existing model set according to the adversarial task; The simulation step size allocation module is used to allocate an initial simulation step size to each of the sub-models; The simulation step size correction module is used to dynamically adjust the initial simulation step size of each sub-model according to the running state of the sub-model during the simulation process, so as to obtain the final simulation step size. The time synchronization and simulation data interpolation mechanism establishment module is used to establish a time synchronization and simulation data interpolation mechanism based on the simulation progress order and causal relationship of all the sub-models, using the smallest final simulation step size among all the sub-models as the reference step size.

[0018] The present invention also discloses a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the above-mentioned method for constructing adversarial simulation models with arbitrary multi-step lengths, so as to solve the technical problems in the prior art, such as the inability to balance the computational efficiency and simulation accuracy of synchronous simulation under fixed step length and the difficulty in guaranteeing the spatiotemporal consistency of simulation models with different step lengths.

[0019] The present invention also discloses a computer-readable storage medium storing a computer program that executes the above-described method for constructing adversarial simulation models with arbitrary multi-step lengths, in order to solve the technical problems in the prior art, such as the inability to balance computational efficiency and simulation accuracy in synchronous simulation with fixed step lengths, and the difficulty in guaranteeing the spatiotemporal consistency of simulation models with different step lengths.

[0020] This invention constructs an adversarial simulation model that balances computational efficiency and simulation accuracy by assigning initial simulation step sizes to different sub-models and dynamically adjusting them to obtain the final simulation step size. At the same time, by designing time synchronization and simulation data interpolation mechanisms through simulation progression order and causal relationship, it can ensure the spatiotemporal consistency of simulation results of sub-models with different simulation step sizes in the adversarial simulation model.

[0021] The beneficial effects of this invention are: 1. The method of the present invention solves the problems of resource management and scheduling loss in fixed step size synchronous simulation propulsion in the prior art, as well as the problems of unbalanced data output and accumulation of accuracy error caused by asynchronous simulation propulsion with fixed hybrid step size scheduling, thereby improving computational efficiency and simulation accuracy; 2. The adversarial simulation model constructed by the method of this invention avoids the problems of data frequency imbalance, low data output frequency due to low simulation resolution, and large prediction deviation caused by data smoothing preprocessing; it also avoids the problem of excessive memory processing burden and impact on the overall computing performance of parallel simulation due to excessively high simulation resolution; and it also avoids the problem of the accumulation of system entity running accuracy error as the simulation resolution is locally too high or too low, reducing the simulation realism and credibility. Attached Figure Description

[0022] Figure 1 This is a flowchart of the multi-step adversarial simulation model construction method disclosed in the embodiments of the present invention; Figure 2 This is a structural block diagram of the multi-step adversarial simulation model construction system disclosed in the embodiments of the present invention; Figure 3 This is a schematic diagram of the computer device disclosed in the embodiments of the present invention.

[0023] Among them, 201 is the model partitioning module; 202 is the simulation step size allocation module; 203 is the simulation step size correction module; 204 is the time synchronization and simulation data interpolation mechanism establishment module; 301 is the memory; and 302 is the processor. Detailed Implementation

[0024] The embodiments of this application will now be described in detail with reference to the accompanying drawings.

[0025] The following specific examples illustrate the implementation of this application. Those skilled in the art can easily understand other advantages and effects of this application from the content disclosed in this specification. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. This application can also be implemented or applied through other different specific embodiments, and the details in this specification can also be modified or changed based on different viewpoints and applications without departing from the spirit of this application. It should be noted that, in the absence of conflict, the following embodiments and features of the embodiments can be combined with each other. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0026] This invention discloses a method for constructing a multi-step adversarial simulation model, see [link to relevant documentation]. Figure 1 As shown, the method includes the following steps: S1. Select multiple sub-models from the existing model set according to the adversarial task, and assign an initial simulation step size to each sub-model; S2. Based on the running status of the sub-models during the simulation process, the initial simulation step size of each sub-model is dynamically adjusted according to the running status to obtain the final simulation step size; S3. Based on the simulation progression order and causal relationship of all the sub-models, the smallest final simulation step size among all the sub-models is used as the reference step size to establish a time synchronization and simulation data interpolation mechanism to obtain the adversarial simulation model.

[0027] Further, in step S1, multiple sub-models are selected from the existing model set according to the adversarial task, and an initial simulation step size is assigned to each sub-model, including: S11. Perform task analysis on the adversarial mission to obtain multiple combat units, and select a sub-model corresponding to each combat unit from the existing model set. For example, for a certain adversarial mission, task analysis can yield combat units including electronic systems, air defense systems, mobile systems, logistics support, radar, missiles, communications, and command, etc., with certain causal and progressive relationships between different combat units.

[0028] S12. Analyze the load change rate for each combat unit and determine the time scale for each sub-model based on the load change rate. In specific implementation, the load change rate of each combat unit is obtained through analysis. Different combat units have different load change rates. For example, electronic system combat units, air defense system combat units, radar combat units, missile combat units, and mobile system combat units all have relatively fast load change rates, and these rates vary. In contrast, logistics support combat units, command combat units, and communication combat units have relatively slow load change rates. When determining the time scale, the time scale for relatively fast load change rates can be set to milliseconds, seconds, or increments, while the time scale for relatively slow load change rates can be set to seconds, minutes, etc.

[0029] S13. Assign an initial simulation step size to each of the sub-models according to the time scale. In specific implementation, the principle of gradually increasing the initial simulation step size as the time scale gradually increases shall be followed.

[0030] Furthermore, in step S13, an initial simulation step size is assigned to each of the sub-models according to the time scale, including: S131. Set a baseline simulation step size corresponding to the baseline time scale; S132. Calculate the multiple relationship between the reference time scale and the time scale of each sub-model; S133. The product of the calculated multiple relationship and the baseline simulation step size is used as the initial simulation step size for each sub-model.

[0031] Specifically, when the reference timescale is in the millisecond range, the reference simulation step size can be set to 10 milliseconds. When assigning the initial simulation step size to each sub-model, the multiple of its timescale to the reference timescale can be calculated. Multiplying the calculated multiple by 10 milliseconds yields the initial simulation step size for that sub-model. For example, for the sub-model corresponding to the electronic system combat unit, whose timescale is in the millisecond range, a simulation step size of 10 milliseconds, the same as the reference simulation step size, can be assigned. For the sub-model corresponding to the logistics support combat unit, whose timescale is in the minute range, which is 600 times the reference timescale, a simulation step size of 600 times the reference simulation step size, i.e., 600 * 10 milliseconds, or 6 seconds, can be assigned as the initial simulation step size. As another example, for the sub-model corresponding to the radar combat unit, whose timescale is in the second range, a simulation step size of 1000 milliseconds, 100 times the reference simulation step size, can be assigned as the initial simulation step size.

[0032] Furthermore, when allocating the initial simulation step size, it is necessary to ensure that the allocated initial simulation step size can meet the simulation accuracy requirements of its sub-model. For example, for the sub-model corresponding to the electronic system combat unit, its simulation accuracy can be set to 1 millisecond to 100 milliseconds, and the initial simulation step size allocated to it also needs to be controlled within this range, and it also needs to be kept within this range during subsequent dynamic adjustments; for the sub-model corresponding to the logistics support combat unit, its simulation accuracy can be set to 1 second to 10 minutes, and the initial simulation step size allocated to it also needs to be controlled within this range, and it also needs to be kept within this range during subsequent dynamic adjustments.

[0033] Further, in step S2, obtaining the running state of the sub-model during the simulation includes: S21. Insert log, callback function, or state variable update code at the key nodes of each sub-model. In specific implementation, log recording code can be inserted at the key nodes of each sub-model to output the running status; callback functions can also be set in the simulation framework to trigger the status check and output the running status after each sub-model has run; or a state variable can be defined for each sub-model to update and store the running status in real time.

[0034] S22. During the simulation, the state information output by each sub-model is acquired in real time, and the running state of each sub-model is obtained by analyzing the state information. The running state includes combat intensity score and error accumulation.

[0035] Further, in step S2, the initial simulation step size of each sub-model is dynamically adjusted according to the running state to obtain the final simulation step size, including: S23. Based on the environmental factors of the confrontation mission and the combat capabilities, personnel status, and resource support of each combat unit in the confrontation mission, calculate the combat intensity range of each combat unit. In specific implementation, the environmental factors of the confrontation mission include terrain conditions (such as mountains, plains, cities) and weather conditions (such as rain, snow, fog, haze, etc.). Combat capabilities include firepower intensity: number of weapons, range, accuracy, lethality, etc.; mobility: speed, mobility range, terrain adaptability, etc.; protection capabilities: armor strength, concealment, fortifications, etc. Personnel status includes troop size: number of personnel in the combat unit; training level: soldiers' training level and combat experience; morale: soldiers' morale and psychological state. Resource support includes ammunition reserves: quantity and type of ammunition; logistical support: supply lines, medical support, maintenance capabilities, etc.

[0036] When calculating the range of combat intensity, weights can be assigned to assessment indicators such as combat capability, personnel status, resource support, and environmental factors. The score of each assessment indicator can be calculated based on the actual situation of each combat unit, and the range of combat intensity for each combat unit can be obtained by weighted averaging.

[0037] S24. Set an error accumulation threshold for each of the sub-models.

[0038] S25. During the simulation, obtain the combat intensity score and error accumulation of each sub-model; S26. When the cumulative error of the sub-model is less than the error accumulation threshold, the initial simulation step size is gradually increased using any one of the following algorithms—variable look-ahead simulation advancement algorithm, range accuracy simulation advancement algorithm, and adaptive algorithm—with the combat intensity score within the combat intensity range as a constraint, to obtain the final simulation step size. In specific implementation, when the cumulative error of a sub-model is less than the error accumulation threshold and its combat intensity score is within the combat intensity range, it indicates that the simulation results of that sub-model are relatively accurate when simulated at its initial simulation step size, but the simulation efficiency may be slightly lower than optimal. Therefore, the initial simulation step size can be increased while ensuring the accuracy of the simulation results to improve simulation efficiency. For example, when the combat intensity score output by the sub-model corresponding to the radar combat unit is within the combat intensity range, its initial simulation step size can be appropriately increased by 5 milliseconds to improve computational efficiency; when the sub-model corresponding to the command combat unit is in a critical decision-making state, its simulation step size can be appropriately reduced to 10 milliseconds to improve simulation accuracy.

[0039] S27. When the cumulative error of the sub-model is greater than or equal to the error accumulation threshold, the initial simulation step size is gradually reduced using any one of the following algorithms: variable look-ahead simulation advancement algorithm, distance accuracy simulation advancement algorithm, and adaptive algorithm, with the combat intensity score within the combat intensity range as a constraint, until a final simulation step size is obtained that satisfies the requirement of cumulative error < error accumulation threshold. In specific implementation, if the cumulative error is greater than the error accumulation threshold, it indicates that the accuracy of the simulation results is not satisfactory when performing simulation based on the initial simulation step size. In this case, the initial simulation step size can be reduced to meet the simulation accuracy requirement of cumulative error < error accumulation threshold, provided that the combat intensity score is within the combat intensity range.

[0040] Furthermore, in steps S26 and S27, the initial simulation step size is dynamically adjusted using either a variable look-ahead simulation approach or a distance-accuracy simulation approach, which allows for a balance between detail capture and computational performance for each sub-model. Specifically, the variable look-ahead simulation approach involves predicting the look-ahead quantity for the next simulation after the current simulation of the sub-model using empirical parameters or statistical methods, and then dynamically adjusting the current simulation step size based on the look-ahead quantity to obtain the simulation step size for the next simulation.

[0041] The distance accuracy simulation advancement method is as follows: obtain the error compensation distance accuracy value of the sub-model, and dynamically adjust the simulation step size of the next simulation after each simulation of the sub-model based on the error compensation distance accuracy value to obtain the simulation step size of the next simulation, wherein the error compensation distance accuracy value is the parameter resolution value of the sub-model in a single cycle.

[0042] Further, in step S3, based on the simulation progression order and causal relationship of all the sub-models, a time synchronization and simulation data interpolation mechanism is established using the smallest final simulation step size among all the sub-models as the reference step size, including: S31. Based on the analysis results of the adversarial task, establish the causal relationship between all the sub-models, and establish a dependency graph based on the causal relationship. In the established dependency graph, a node represents a sub-model, and the edge between two nodes represents the causal relationship between two sub-models. Based on the causal relationship, the order of two sub-models can be known.

[0043] S32. Based on the dependency graph, determine the simulation progression order of each sub-model according to the topology sorting algorithm. For example, when performing simulation calculations, if the output of sub-model A is the input of sub-model B, then sub-model A must complete the simulation before sub-model B.

[0044] S33. According to the simulation progress order, the smallest final simulation step size among all the sub-models is used as the reference step size. The simulation time is divided into multiple synchronization points. For the sub-models whose final simulation step size is greater than the reference step size, data interpolation is performed at each synchronization point to establish a time synchronization and simulation data interpolation mechanism.

[0045] In practical implementation, for sub-models with relatively large final simulation step sizes, the sub-model with the smallest final simulation step size can be used as the benchmark to maintain temporal synchronization and avoid spatiotemporal inconsistencies. For example, in the time synchronization and simulation data interpolation mechanism, asynchronous time management lists can be used for time synchronization. These lists specify the order and causal relationships of the simulation progression for each sub-model, preventing disruption of the causal relationships in the simulation process. When outputting simulation results, high-frequency output data can be generated through data interpolation (including linear interpolation or higher-order interpolation) to maintain consistency with the output data of the sub-model with the smallest final simulation step size.

[0046] In an improved embodiment of the above-described multi-step adversarial simulation model construction method, the method further includes: S4. A distributed parallel computing architecture is adopted, and computing nodes are allocated to each sub-model in the adversarial simulation model for parallel simulation calculation to improve the overall computing efficiency.

[0047] In an improved embodiment of the above-described multi-step adversarial simulation model construction method, the method further includes: S5. Evaluate and optimize the constructed adversarial simulation model. In specific implementation, training samples can be established through completed adversarial tasks. The training samples are input into the adversarial simulation model for simulation training. The output simulation training results are compared and analyzed with the real results of the adversarial task. Based on the analysis results, the final simulation step size of each sub-model in the constructed adversarial simulation model is optimized.

[0048] Based on the same inventive concept, this invention also provides a multi-step adversarial simulation model construction system, as described in the following embodiments. Since the principle of the multi-step adversarial simulation model construction system in solving the problem is similar to the multi-step adversarial simulation model construction method disclosed in the above embodiments, the implementation of the adversarial simulation model construction system can refer to the implementation of the above adversarial simulation model construction method, and repeated details will not be elaborated further. The terms "unit" or "module" refer to a combination of software and / or hardware capable of achieving a predetermined function. Although the apparatus described in the following embodiments is preferably implemented in software, hardware implementation, or a combination of software and hardware, is also possible and contemplated.

[0049] Figure 2 This is a structural block diagram of a multi-step adversarial simulation model construction system disclosed in an embodiment of the present invention, such as... Figure 2 As shown, the adversarial simulation model construction system includes: a model partitioning module 201, a simulation step size allocation module 202, a simulation step size correction module 203, and a time synchronization and simulation data interpolation mechanism establishment module 204. The structure is described below.

[0050] The model partitioning module 201 is used to select multiple sub-models from the existing model set according to the adversarial task; The simulation step size allocation module 202 is used to allocate an initial simulation step size to each of the sub-models; The simulation step size correction module 203 is used to obtain the running status of the sub-model during the simulation process, and dynamically adjust the initial simulation step size of each sub-model according to the running status to obtain the final simulation step size. The time synchronization and simulation data interpolation mechanism establishment module 204 is used to establish a time synchronization and simulation data interpolation mechanism based on the simulation progress order and causal relationship of all the sub-models, using the smallest final simulation step size among all the sub-models as the reference step size.

[0051] This invention constructs an adversarial simulation model that balances computational efficiency and simulation accuracy by assigning initial simulation step sizes to different sub-models and dynamically adjusting them to obtain the final simulation step size. Furthermore, by designing time synchronization and simulation data interpolation mechanisms based on simulation progression order and causal relationships, the spatiotemporal consistency of simulation results for sub-models with different simulation step sizes within the adversarial simulation model can be ensured. Compared with existing technologies, the beneficial effects achieved by at least one of the above-mentioned technical solutions adopted in the embodiments of this specification include at least the following: 1. The method of the present invention solves the problems of resource management and scheduling loss in fixed step size synchronous simulation propulsion in the prior art, as well as the problems of unbalanced data output and accumulation of accuracy error caused by asynchronous simulation propulsion with fixed hybrid step size scheduling, thereby improving computational efficiency and simulation accuracy; 2. The adversarial simulation model constructed by the method of this invention avoids the problems of data frequency imbalance, low data output frequency due to low simulation resolution, and large prediction deviation caused by data smoothing preprocessing; it also avoids the problem of excessive memory processing burden and impact on the overall computing performance of parallel simulation due to excessively high simulation resolution; and it also avoids the problem of the accumulation of system entity running accuracy error as the simulation resolution is locally too high or too low, reducing the simulation realism and credibility.

[0052] In this embodiment, a computer device is provided, such as... Figure 3 As shown, it includes a memory 301, a processor 302, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the above-described method for constructing adversarial simulation models with arbitrary multi-step lengths.

[0053] Specifically, the computer device can be a computer terminal, a server, or a similar computing device.

[0054] In this embodiment, a computer-readable storage medium is provided, which stores a computer program that executes the above-described method for constructing adversarial simulation models of any multi-step length.

[0055] Specifically, computer-readable storage media include both permanent and non-permanent, removable and non-removable media, which can store information using any method or technology. Information can be computer-readable instructions, data structures, program modules, or other data. Examples of computer-readable storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, CD-ROM, digital versatile optical disc (DVD) or other optical storage, magnetic tape, magnetic disk storage or other magnetic storage devices, or any other non-transferable medium that can be used to store information accessible by a computing device. As defined herein, computer-readable storage media do not include transient media, such as modulated data signals and carrier waves.

[0056] Obviously, those skilled in the art should understand that the modules or steps of the above-described embodiments of the present invention can be implemented using general-purpose computing devices. They can be centralized on a single computing device or distributed across a network of multiple computing devices. Optionally, they can be implemented using computer-executable program code, thereby storing them in a storage device for execution by a computing device. In some cases, the steps shown or described can be performed in a different order than those presented here, or they can be fabricated as separate integrated circuit modules, or multiple modules or steps can be fabricated as a single integrated circuit module. Thus, the embodiments of the present invention are not limited to any particular hardware and software combination.

[0057] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, various modifications and variations can be made to the embodiments of the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for constructing a multi-step adversarial simulation model, characterized in that, The method includes the following steps: S1. Select multiple sub-models from the existing model set according to the adversarial task, and assign an initial simulation step size to each sub-model; S2. Based on the running status of the sub-models during the simulation process, the initial simulation step size of each sub-model is dynamically adjusted according to the running status to obtain the final simulation step size; S3. Based on the simulation progression order and causal relationship of all the sub-models, the smallest final simulation step size among all the sub-models is used as the reference step size to establish a time synchronization and simulation data interpolation mechanism to obtain the adversarial simulation model.

2. The method for constructing a multi-step adversarial simulation model according to claim 1, characterized in that, In step S1, multiple sub-models are selected from the existing model set according to the adversarial task, and an initial simulation step size is assigned to each sub-model, including: S11. Perform task analysis on the adversarial task to obtain multiple combat units, and select the sub-model corresponding to each combat unit from the existing model set; S12. Analyze and obtain the load change rate for each combat unit, and determine the time scale of each sub-model based on the load change rate; S13. Assign an initial simulation step size to each of the sub-models according to the time scale.

3. The method for constructing a multi-step adversarial simulation model according to claim 2, characterized in that, In step S13, an initial simulation step size is assigned to each sub-model according to the time scale, including: S131. Set a baseline simulation step size corresponding to the baseline time scale; S132. Calculate the multiple relationship between the reference time scale and the time scale of each sub-model; S133. The product of the calculated multiple relationship and the baseline simulation step size is used as the initial simulation step size for each sub-model.

4. The method for constructing a multi-step adversarial simulation model according to claim 1, characterized in that, In step S2, the running state of the sub-model during the simulation is obtained, including: S21. Insert logs, callback functions, or state variable update code at the key nodes of each sub-model; S22. During the simulation, the state information output by each sub-model is acquired in real time, and the running state of each sub-model is obtained by analyzing the state information. The running state includes combat intensity score and error accumulation.

5. The method for constructing a multi-step adversarial simulation model according to claim 1, characterized in that, In step S2, the initial simulation step size of each sub-model is dynamically adjusted according to the running state to obtain the final simulation step size, including: S23. Calculate the combat intensity range of each combat unit based on the environmental factors of the confrontation mission and the combat capability, personnel status and resource support of each combat unit in the confrontation mission. S24. Set an error accumulation threshold for each of the sub-models; S25. During the simulation, obtain the combat intensity score and error accumulation of each sub-model; S26. When the cumulative error of the sub-model is less than the cumulative error threshold, the combat intensity score is used as a constraint within the combat intensity range. The initial simulation step size is gradually increased by any one of the variable look-ahead simulation advancement algorithm, distance accuracy simulation advancement algorithm, and adaptive algorithm to obtain the final simulation step size. S27. When the cumulative error of the sub-model is greater than or equal to the error accumulation threshold, the initial simulation step size is gradually reduced by any one of the variable look-ahead simulation advancement algorithm, distance accuracy simulation advancement algorithm, and adaptive algorithm, with the combat intensity score within the combat intensity range as a constraint, until the final simulation step size satisfies the condition that the cumulative error is less than the error accumulation threshold.

6. The method for constructing a multi-step adversarial simulation model according to claim 5, characterized in that, In steps S26 and S27: the initial simulation step size is dynamically adjusted using either a variable look-ahead simulation propulsion method or a distance accuracy simulation propulsion method; The variable look-ahead simulation advancement method is as follows: predict the look-ahead of the next simulation after the current simulation of the sub-model by using empirical parameters or statistical methods, and dynamically adjust the current simulation step size corresponding to the current simulation based on the look-ahead, so as to obtain the simulation step size of the next simulation. The distance accuracy simulation advancement method is as follows: obtain the error compensation distance accuracy value of the sub-model, and dynamically adjust the simulation step size of the next simulation after each simulation of the sub-model based on the error compensation distance accuracy value to obtain the simulation step size of the next simulation, wherein the error compensation distance accuracy value is the parameter resolution value of the sub-model in a single cycle.

7. The method for constructing a multi-step adversarial simulation model according to claim 1, characterized in that, In step S3, based on the simulation progression order and causal relationships of all the sub-models, and using the smallest final simulation step size among all the sub-models as the baseline step size, a time synchronization and simulation data interpolation mechanism is established, including: S31. Based on the analysis results of the adversarial task, establish the causal relationship between all the sub-models, and establish a dependency graph based on the causal relationship. In the established dependency graph, a node represents a sub-model, and the edge between two nodes represents the causal relationship between two sub-models. Based on the causal relationship, the order of two sub-models can be known. S32. Based on the dependency graph, determine the simulation progression order of each sub-model according to the topology sorting algorithm; S33. According to the simulation progress order, the smallest final simulation step size among all the sub-models is used as the reference step size. The simulation time is divided into multiple synchronization points. For the sub-models whose final simulation step size is greater than the reference step size, data interpolation is performed at each synchronization point to establish a time synchronization and simulation data interpolation mechanism.

8. The method for constructing a multi-step adversarial simulation model according to claim 1, characterized in that, The method further includes: S4. A distributed parallel computing architecture is adopted, and computing nodes are allocated to each sub-model in the adversarial simulation model for parallel simulation computing.

9. A multi-step adversarial simulation model construction system, characterized in that, include: The module includes a model partitioning module, a simulation step size allocation module, a simulation step size correction module, and a time synchronization and simulation data interpolation mechanism establishment module. The model partitioning module is used to select multiple sub-models from the existing model set according to the adversarial task; The simulation step size allocation module is used to allocate an initial simulation step size to each of the sub-models; The simulation step size correction module is used to obtain the running status of the sub-model during the simulation process, and dynamically adjust the initial simulation step size of each sub-model according to the running status to obtain the final simulation step size. The time synchronization and simulation data interpolation mechanism establishment module is used to establish a time synchronization and simulation data interpolation mechanism based on the simulation progress order and causal relationship of all the sub-models, using the smallest final simulation step size among all the sub-models as the reference step size.

10. A computer device, characterized in that, The method includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the multi-step adversarial simulation model construction method according to any one of claims 1-8.

Images