Action plan generation model construction method based on object relation decomposition

Abstract

The invention discloses an action scheme generation model construction method based on object relation decomposition, particularly relates to the field of simulated confrontation, and is used for solving the cross-level constraint conflict problem of action parameters of a high-level instruction and a bottom-level unit when multi-level execution units of two confrontation parties cooperatively execute a composite task. The method comprises the following steps: analyzing internal association between a high-level instruction and an action parameter of a bottom-level unit through object relationship decomposition, constructing a multi-level object relationship graph, and providing support for cross-level collaboration; the dynamic constraint conduction chain design is combined with quantification of influence weights, so that a bottom layer unit can quickly respond to high-level instruction changes, resource configuration is optimized, and timeliness and efficiency of collaborative execution are greatly improved; a conflict analysis technology of fractal aggregation and transition probability evaluation is adopted, cross-level potential conflicts are deeply mined and eliminated, and the stability and reliability of an action scheme are enhanced; in addition, the target expectation of the high-level instruction is dynamically verified through real-time simulation, and it is ensured that the action scheme is highly matched with the intention.

Description

Technical Field

[0001] The present invention relates to the field of simulated confrontation, and more specifically, to a method for constructing an action plan generation model based on object relationship decomposition. Background Art

[0002] The existing simulation system adopts a hierarchical decoupled processing mode, and the deduction layer and the execution layer generate solutions independently, resulting in unpredictable conflict nodes when the high-level instructions are transmitted to the bottom layer. When the command layer adjusts the overall combat rhythm, the tactical unit is forced to repeatedly correct the action parameters within the limited solution space, which not only greatly prolongs the deduction cycle, but also may generate unfeasible solutions due to the failure to resolve the key node conflicts. The root cause of the failure of nested conflict resolution of multi-granularity spatiotemporal constraints is that the dynamic transmission and adaptive coordination model of cross-level constraints has not been established, which causes a significant deviation between the virtual confrontation deduction results and the actual tactical execution effect.

[0003] In order to solve the above problems, a technical solution is now provided. Summary of the invention

[0004] In order to overcome the above-mentioned defects of the prior art, an embodiment of the present invention provides a method for constructing an action plan generation model based on object relationship decomposition, which analyzes the intrinsic relationship between high-level instructions and low-level unit action parameters through object relationship decomposition, constructs a multi-level object relationship graph, and provides support for cross-level collaboration; dynamic constraint transmission chain design, combined with the quantification of influence weights, enables low-level units to respond quickly to changes in high-level instructions, while optimizing resource allocation, greatly improving the timeliness and efficiency of collaborative execution; adopts fractal aggregation and transition probability evaluation conflict analysis technology to deeply explore and eliminate cross-level potential conflicts, and enhance the stability and reliability of action plans; in addition, the target expectations of high-level instructions are dynamically verified through real-time simulation to ensure that the action plan is highly consistent with the intention, so as to solve the problems raised in the above-mentioned background technology.

[0005] To achieve the above object, the present invention provides the following technical solutions: S1. Through object relationship decomposition technology, the spatiotemporal constraints of high-level instructions and the action parameters of low-level units are separated to construct a multi-level object relationship graph; S2. According to the multi-level object relationship graph, traverse the impact path of high-level instruction changes and extract the affected parameter set as the dynamic constraint transmission chain; S3. Sort the parameter sets in the dynamic constraint transmission chain by impact weights and redistribute the action resources of the underlying units to generate a preliminary coordination plan; S4. Scan the cross-level conflicts in the preliminary coordination plan, identify and evaluate the severity of the cross-level conflicts in the preliminary coordination plan, and use iterative optimization methods to adjust parameters to effectively resolve the conflicts; S5. Integrate the adjusted parameter set back into the object relationship graph, verify the expected goals of the high-level instructions through real-time simulation, and generate the final action plan.

[0006] In a preferred embodiment, step S1 includes the following: First, parse the high-level instructions to extract their time constraints and space constraints. The time constraint is defined as the time interval from the start time to the deadline of the task, and the space constraint is defined as the coordinate set of the task area. Subsequently, identify the action parameters of each underlying unit. Then, model the high-level instructions as nodes containing time constraints and space constraints, and model the underlying units as nodes containing a set of action parameters. Introduce intermediate constraint nodes to represent the sub-constraints after the decomposition of the high-level instruction constraints. Connect the high-level instruction nodes to the intermediate constraint nodes and the intermediate constraint nodes to the underlying unit nodes in sequence through directed edges to reflect the constraint decomposition and the restrictive relationship on the underlying units. Decompose the time constraint of the high-level instruction into multiple sub-time periods and the space constraint into multiple sub-regions, and calculate the influence intensity weights of each sub-constraint on the underlying units. The weight calculation method is the product of the length of the sub-time period and the area of the sub-region divided by the product of the total length of the high-level instruction time interval and the total area of the space region. Finally, form a multi-level object relationship graph structure containing nodes and their weighted directed edges.

[0007] In a preferred embodiment, step S2 includes the following: First, identify the changes by calculating the change amounts of the time constraint and space constraint of the high-level instructions. The time change amount is defined as the sum of the offsets of the start time and deadline of the changed time interval relative to the original time interval, and the space change amount is defined as the sum of the areas of the non-overlapping parts of the changed space region and the original space region to quantify the adjustment amplitude. Then, use the breadth-first search algorithm to traverse the influence paths from the high-level instruction nodes through the intermediate constraint nodes to the underlying unit nodes, and record each complete path. Next, for the intermediate constraint nodes on each path, calculate the time change amount and space change amount of their sub-constraints, and compare them with the preset time influence threshold and space influence threshold respectively. If any change amount exceeds the corresponding threshold, mark the action parameters of the underlying unit corresponding to this path as affected parameters. Finally, organize all the sets of affected action parameters into dynamic constraint conduction chains. Each chain includes the high-level instruction node, the affected intermediate constraint node, and the affected underlying unit node and its action parameters to reflect the transmission path of the influence.

[0008] In a preferred embodiment, step S3 includes the following: First, calculate the influence weight for each affected action parameter in the dynamic constraint conduction chain. The influence weight is determined by comprehensively considering the time influence factor, space influence factor, and resource shortage degree. The time influence factor is calculated by dividing the sub-time change amount of the intermediate constraint node by the length of the time interval after the change of the high-level instruction. The space influence factor is calculated by dividing the sub-space change amount of the intermediate constraint node by the area of the space region after the change of the high-level instruction. The resource shortage degree is calculated by dividing the current resource usage amount of the bottom-level unit by the total resource amount. Then, take the square root of the sum of the squares of the time influence factor and the space influence factor, and multiply it by the resource shortage adjustment term to obtain the influence weight. Then, sort the set of affected action parameters in descending order according to the influence weight to form a priority processing list. Next, calculate the resource adjustment factor according to the sorting result. The resource adjustment factor is obtained by multiplying the proportion of the influence weight of each affected action parameter by the remaining proportion of the bottom-level unit's resources, and update the resource allocation by multiplying the original resource allocation by the adjustment coefficient, that is, 1 plus the resource adjustment factor, while ensuring that the adjusted resource allocation does not exceed the total resource amount. Finally, update the affected action parameters according to the adjusted resource allocation, and determine the specific update method according to the parameter type such as speed or monitoring radius, and generate a preliminary coordination plan containing all the adjusted affected action parameters.

[0009] In a preferred embodiment, step S4 includes the following content: S4.1, in the processing of conflict identification and spatio-temporal mapping, first map the action parameters of the bottom-level units in the preliminary coordination plan to the spatio-temporal coordinate system to identify the regions inconsistent with the spatio-temporal constraints of the high-level instructions.

[0010] In a preferred embodiment, S4.2, when calculating the fractal aggregation index, the box-counting method is used to quantify the aggregation degree of conflicts in the spatio-temporal grid; the specific calculation logic is as follows: divide the spatio-temporal grid into multiple boxes at different scales, and count the number of boxes containing conflict grid cells at each scale; then perform a regression analysis on the logarithmic relationship between the number of boxes at different scales and the scale, and calculate the fractal dimension by fitting. The fractal dimension reflects the complexity and self-similarity of the conflict grid; then, divide the fractal dimension by the embedding dimension of the spatio-temporal grid, and the embedding dimension is the dimension number of the spatio-temporal grid, so as to obtain the fractal aggregation index.

[0011] In a preferred embodiment, in S4.3, when calculating the conflict transfer index, first construct a conflict transfer probability matrix between the underlying units to quantify the propagation trend of conflicts between the underlying units; the specific calculation logic is as follows: measure the spatial proximity by calculating the average spatial distance of the action trajectories of two underlying units; measure the similarity of action parameters by calculating the difference degree of the action parameters of two underlying units; then use the exponential decay function to calculate the transfer probability and normalize all transfer probabilities to ensure that the sum of the transfer probabilities of each underlying unit is 1; finally, take the maximum eigenvalue of the transfer probability matrix as the conflict transfer index.

[0012] In a preferred embodiment, in S4.4, in the comprehensive conflict severity assessment, combine the fractal aggregation index and the conflict transfer index by weighted summation to calculate the comprehensive conflict severity coefficient.

[0013] In a preferred embodiment, in S4.4, when iteratively optimizing and adjusting parameters, use the gradient descent method to optimize the comprehensive conflict severity coefficient, with the goal of adjusting the action parameters of the underlying units to reduce the conflict severity; the specific calculation logic is as follows: use the action parameters in the preliminary coordination plan as the initial values; calculate the partial derivatives of the comprehensive conflict severity coefficient with respect to each action parameter, and update the action parameters according to the partial derivatives and the learning rate, where the learning rate controls the step size of each adjustment; repeat this process until the comprehensive conflict severity coefficient converges or reaches the preset number of iterations; at the same time, ensure that the adjusted action parameters are within the preset feasible range to avoid exceeding the actual limits, and gradually reduce the conflict severity through iterative optimization, and finally obtain an optimized set of action parameters to minimize the comprehensive conflict severity coefficient.

[0014] In a preferred embodiment, step S5 includes the following contents: First, map the optimized set of action parameters to the underlying unit nodes of the multi-level object relationship graph, and maintain data consistency by updating the action parameter attributes of each underlying unit node to the corresponding optimized values. Then, construct a virtual spatio-temporal environment to real-time simulate and verify the spatio-temporal coverage requirements of high-level instructions. Specifically, by simulating the time interval and spatial region of high-level instructions, calculate the action trajectories of underlying units, and determine whether they meet the coverage requirements of high-level instructions at each time and spatial position. Next, calculate the coverage rate, which is the ratio of the number of spatio-temporal points satisfied by the action trajectories of underlying units to the total number of spatio-temporal points that need to be covered by high-level instructions, and evaluate according to a preset coverage rate threshold. If the coverage rate does not meet the standard, identify the uncovered areas, screen the underlying units that are spatio-temporally adjacent, adjust their key action parameters to expand the coverage range, and repeat the simulation verification until the coverage rate reaches the threshold. Finally, integrate the adjusted set of action parameters, generate the final action parameters and trajectories of all underlying units, ensure that the spatio-temporal coverage requirements of high-level instructions are fully met, and package them into a final action plan that can be used for simulation confrontation.

[0015] Technical effects and advantages of the method for constructing an action plan generation model based on object relationship decomposition of the present invention: Through the object relationship decomposition technology, the present invention deeply analyzes the internal relationship between high-level instructions and the action parameters of underlying units, constructs a multi-level object relationship graph, and provides a solid support for cross-level collaboration; the dynamic constraint conduction chain design, combined with the precise quantification of influence weights, enables the underlying units to quickly respond to changes in high-level instructions, while optimizing resource allocation, and greatly improves the timeliness and efficiency of collaborative execution; the conflict analysis technology using fractal aggregation and transfer probability evaluation deeply mines and resolves potential cross-level conflicts, significantly enhancing the stability and reliability of the action plan; in addition, through real-time simulation technology, dynamically verify the target expectations of high-level instructions to ensure that the action plan highly conforms to the intention; through the organic integration of these technical elements, overcome the bottleneck of cross-level constraint conflicts in simulation confrontation, provide a strong technical guarantee for the efficient collaboration of multi-level execution units on both sides of the simulation, and comprehensively improve the deduction accuracy and efficiency. Description of the Drawings

[0016] Figure 1 It is a structural schematic diagram of the method for constructing an action plan generation model based on object relationship decomposition of the present invention. Detailed Embodiments

[0017] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0018] Example 1: Figure 1 A method for constructing an action plan generation model based on object relationship decomposition according to the present invention is provided, including: S1. Through object relationship decomposition technology, strip the spatio-temporal constraints of high-level instructions from the action parameters of low-level units, and construct a multi-level object relationship map.

[0019] S2. According to the multi-level object relationship map, traverse the influence path of high-level instruction changes, and extract the set of affected parameters as the dynamic constraint conduction chain.

[0020] S3. For the set of parameters in the dynamic constraint conduction chain, sort them according to the influence weight and reallocate the action resources of low-level units to generate a preliminary coordination plan.

[0021] S4. Scan the cross-level conflicts in the preliminary coordination plan, identify and evaluate the severity of the cross-level conflicts in the preliminary coordination plan, and use an iterative optimization method to adjust the parameters to effectively resolve the conflicts.

[0022] S5. Integrate the adjusted set of parameters back into the object relationship map, and verify the expected goals of high-level instructions through real-time simulation to generate the final action plan.

[0023] In a simulated confrontation scenario, high-level instructions usually contain complex spatio-temporal constraints, such as the time window and spatial coverage of tasks, while the action parameters of low-level units (such as drones, ships, etc.) (such as speed, heading, etc.) need to be coordinated according to these constraints. The goal of step S1 is to strip the spatio-temporal constraints of high-level instructions from the action parameters of low-level units through object relationship decomposition technology, and construct a multi-level object relationship map to provide a structured basis for the impact analysis and resource allocation of high-level instruction changes.

[0024] Step S1 includes the following: S1.1. During the extraction of spatio-temporal constraints of high-level instructions, first parse the high-level instructions to separate the time constraints and spatial constraints contained therein. The extraction of time constraints is to identify the task start time and task end time described in the high-level instructions respectively, and calculate the difference between these two time points as a time interval, in seconds, representing the duration of task execution. The extraction of spatial constraints is to identify the task area described in the high-level instructions as a geometric shape, such as a rectangular area, and record the set of boundary coordinates of this area to represent the spatial range required to cover the task. Through this parsing and quantification, the abstract description of high-level instructions is transformed into specific time and spatial parameters, providing a data basis for subsequent decomposition. The final output is the time constraint parameter, that is, the time interval from the task start time to the end time, and the spatial constraint parameter, that is, the set of coordinates of the task area.

[0025] S1.2. During the identification process of the action parameters of the underlying units, each underlying unit is analyzed independently to determine its specific action characteristics. Each underlying unit is first identified by its unique number to ensure the accuracy of subsequent associations. Then, the action parameters of each underlying unit are extracted, including speed, heading, and monitoring radius, etc. Among them, speed represents the distance moved per unit time, with the unit of meters per second; heading represents the direction of movement, with the unit of radians; monitoring radius represents the size of the perception range, with the unit of meters. By identifying and recording these parameters one by one, a set of descriptions of the action capabilities of the underlying units is established, laying a foundation for the matching with the high-level instruction constraints. The final output is the set of action parameters for each underlying unit, containing the quantified values of all relevant parameters.

[0026] S1.3. During the decomposition process of the object relationships, the high-level instructions and the underlying units are respectively modeled as nodes in a graph, and the connecting edges are established through the constraint relationships. First, the high-level instructions are modeled as a node, including their time constraints and space constraint attributes; each underlying unit is modeled as a node, including its set of action parameters. Secondly, intermediate constraint nodes are introduced to represent the sub-constraints formed after the decomposition of the high-level instruction constraints. Then, directed edges are established. Among them, the edge from the high-level instruction node to the intermediate constraint node represents the process of decomposing the time constraints and space constraints into several sub-parts; the edge from the intermediate constraint node to the underlying unit node represents the specific restrictions of these sub-constraints on the action parameters of the underlying units. Through the structured modeling of nodes and edges, the overall constraints of the high-level instructions are decomposed into assignable sub-constraints, and their association relationships with the underlying units are clarified. Finally, a graph structure containing high-level instruction nodes, intermediate constraint nodes, underlying unit nodes, and their connecting edges is formed.

[0027] S1.4. During the construction of the multi-level object relationship graph, a three-layer structure is formed based on the decomposition result. The first layer is the high-level instruction node, which contains the complete task time constraint and space constraint; the second layer is the intermediate constraint node, representing the decomposed sub-constraints; the third layer is the low-level unit node, which contains the respective action parameter sets. The decomposition of the time constraint is to divide the time interval of the high-level instruction into multiple sub-time periods, and the sum of these sub-time periods covers the original time interval; the decomposition of the space constraint is to divide the space area of the high-level instruction into multiple sub-areas, and the sum of these sub-areas covers the original space area. The influence intensity of the sub-constraint on the low-level unit is calculated as follows: multiply the length of each sub-time period by the area of the corresponding sub-area to obtain a product value, and then divide this product value by the product of the total length of the high-level instruction time interval and the total area of the space area to obtain the weight of the influence intensity. Through the hierarchical and weighted method, the constraints are decomposed step by step and the influence on the low-level unit is quantified. At the same time, a timestamp is introduced into the graph to record the update time after the change of the high-level instruction to support dynamic adjustment. Finally, a multi-level object relationship graph containing three layers of nodes and weighted directed edges is constructed.

[0028] For step S2, the directed edges in the graph clarify the influence path from the high-level instruction node through the intermediate constraint node to the low-level unit node. By traversing along these edges in step S2, the set of action parameters of the low-level units affected by the change of the high-level instruction can be directly extracted. For step S3, the action parameters affected by the constraint and their corresponding influence intensity weights are marked in the graph. According to these weight information in step S3, a dynamic constraint conduction chain from the change of the high-level instruction to the influence on the low-level unit is constructed and sorted according to the weight size. Through the structured and quantified data of the graph, it is ensured that the subsequent steps can efficiently utilize the decomposition result of step S1 to achieve the continuity of influence analysis and resource allocation.

[0029] Through the above processing technical logic, step S1 successfully separates the time constraint and space constraint of the high-level instruction from the action parameters of the low-level unit and constructs a multi-level object relationship graph. Through the hierarchical structure of the high-level instruction node, intermediate constraint node and low-level unit node, as well as the weighted directed edges connecting these nodes, the decomposition and transmission relationship of the constraint is clearly expressed. This structured data provides precise support for the subsequent steps, enabling step S2 to traverse the influence path and step S3 to extract the dynamic constraint conduction chain and perform sorting analysis.

[0030] In step S1, through the object relationship decomposition technology, the spatio-temporal constraints of the high-level instructions and the action parameters of the underlying units are stripped, and a multi-level object relationship map is constructed. This map, in the form of high-level instruction nodes, intermediate constraint nodes, and underlying unit nodes, shows how high-level instructions affect the action parameters of underlying units through the decomposed sub-constraints. The purpose of step S2 is to traverse the influence path of high-level instruction changes according to this map, extract the set of affected underlying unit action parameters, and form a dynamic constraint conduction chain, providing an accurate basis for resource allocation in step S3 and conflict resolution in step S4.

[0031] Step S2 includes the following: S2.1, Identification of high-level instruction changes: In the process of identifying high-level instruction changes, first, the time constraints and space constraints of the original high-level instructions and the changed high-level instructions are compared. The change in time constraints is determined by calculating the offset of the task start time and end time. The specific method is to add the absolute value of the difference between the changed start time and the original start time to the absolute value of the difference between the changed end time and the original end time to obtain the total amount of time change. The change in space constraints is measured by calculating the sum of the areas of the non-overlapping parts between the changed space region and the original space region. The non-overlapping parts include the parts of the changed space region that do not belong to the original space region, and the parts of the original space region that do not belong to the changed space region. By quantifying the degree of change in both the time and space dimensions, the amplitude of high-level instruction changes is accurately evaluated, providing a basic basis for subsequent analysis of the influence path. The final result is the specific values of the time change amount and space change amount of the high-level instructions.

[0032] S2.2, In the process of traversing the influence path, the breadth-first search method is adopted. Starting from the high-level instruction node, visit the intermediate constraint nodes connected to it layer by layer, and then further visit the underlying unit nodes connected to the intermediate constraint nodes. Breadth-first search ensures that each node is visited along the shortest path and avoids repeated visits to nodes that have already been processed. During this process, record each complete path starting from the high-level instruction node, passing through the intermediate constraint nodes, and finally reaching the underlying unit node. Each path corresponds to an underlying unit and its related action parameters. Through a systematic graph traversal method, all underlying units and their action parameters that may be affected by high-level instruction changes are comprehensively identified, ensuring that no affected part is missed. The final result is a set containing all affected paths, and each path clearly lists the high-level instruction node, intermediate constraint node, and underlying unit node.

[0033] S2.3. During the extraction of affected parameters, each traversed path is analyzed, with a focus on examining the change amount of the sub-constraints corresponding to the intermediate constraint nodes on the path. The change amount of sub-constraints includes the time change amount and the space change amount. The time change amount reflects the degree of change in the sub-time period, and the space change amount reflects the degree of change in the sub-region. Set a time impact threshold and a space impact threshold. If the sub-constraint time change amount of a certain intermediate constraint node exceeds the time impact threshold, or the sub-constraint space change amount exceeds the space impact threshold, it is determined that the intermediate constraint node is affected, and then the underlying unit action parameter corresponding to this path is marked as an affected parameter. By screening through setting thresholds, the part where the high-level instruction change has a significant impact on the underlying unit action parameter is accurately extracted, avoiding including irrelevant parameters in the analysis scope. The final result is a set of affected underlying unit action parameters.

[0034] S2.4. During the construction of the dynamic constraint conduction chain, the set of extracted affected action parameters is organized into a chain structure. The starting point of each chain is the high-level instruction node, the middle part is the affected intermediate constraint nodes, and the end point is the affected underlying unit node and its corresponding action parameters. This chain structure reflects the specific impact path by which the high-level instruction change is gradually transmitted to the underlying unit action parameter through the intermediate constraint nodes. Through the chain representation method, the transmission order and hierarchical relationship of the impact are intuitively displayed, providing clear structured data support for subsequent resource allocation and conflict resolution. The final result is a set of dynamic constraint conduction chains, and each chain contains the high-level instruction node, the affected intermediate constraint nodes, and the affected underlying unit node and its action parameters.

[0035] Step S3 uses the set of affected action parameters in the set of dynamic constraint conduction chains, sorts them according to the impact weight, and reallocates the action resources of the underlying units accordingly. At the same time, Step S4 scans the cross-level conflicts in the preliminary coordination plan based on the affected action parameters in the set of dynamic constraint conduction chains and eliminates the conflicts by adjusting these parameters. By providing accurate affected action parameters and their impact paths, it is ensured that the subsequent steps can directly utilize the analysis results of the current step, realizing the continuity and efficiency of resource allocation and conflict resolution. The final result is the set of dynamic constraint conduction chains as a data bridge connecting the processing flows of the current step and the subsequent steps.

[0036] The goal of Step S3 is to process the set of affected parameters in the dynamic constraint conduction chain, generate a preliminary coordination plan by sorting according to the impact weight and reallocating the action resources of the underlying units, laying a foundation for the conflict detection and resolution in Step S4.

[0037] Step S3 includes the following: S3.1. During the process of calculating the influence weight, first define an influence weight value for each affected action parameter in the dynamic constraint conduction chain. This influence weight is jointly determined by three parts: the time influence factor, the space influence factor, and the resource shortage degree. The calculation method of the time influence factor is to divide the sub-time change amount of the intermediate constraint node by the time interval length after the change of the high-level instruction. This result reflects the influence degree of the time constraint change of the high-level instruction on this affected action parameter. The calculation method of the space influence factor is to divide the sub-space change amount of the intermediate constraint node by the space area after the change of the high-level instruction. This result reflects the influence degree of the space constraint change of the high-level instruction on this affected action parameter. The calculation method of the resource shortage degree is to divide the current resource usage amount of the bottom layer unit by the total amount of resources. This result reflects the tightness of the resource usage of the bottom layer unit. The final calculation of the influence weight adopts a non-linear combination method: first add the squares of the time influence factor and the space influence factor and then take the square root to obtain a comprehensive influence value, and then multiply this comprehensive influence value by a regulation term, which is composed of 1 plus the resource shortage degree, so as to enhance the regulation effect of the resource shortage degree on the influence weight. By quantifying the effect of high-level instruction changes on the affected action parameters of the bottom layer unit in multiple dimensions and combining the resource shortage situation, ensure that the influence weight can accurately reflect the adjustment priority. The final result obtained is the influence weight value of each affected action parameter.

[0038] S3.2. During the calculation process of sorting the affected parameters, according to the influence weight calculated above, sort the set of affected action parameters in the dynamic constraint conduction chain in descending order. The specific method is to sort all affected action parameters from high to low according to their influence weight values. The affected action parameters with higher influence weight values are arranged in the front of the list, and the affected action parameters with lower influence weight values are arranged in the back of the list. This sorting method ensures that during subsequent resource allocation, the affected action parameters that respond more strongly to high-level instruction changes and are resource-scarce are processed first. By quantifying and sorting the influence weight, reasonably determine the priority order of resource allocation, so as to improve the resource utilization efficiency and ensure timely response to the changes of high-level instructions. The final result obtained is a list of affected action parameters sorted in descending order of influence weight values.

[0039] S3.3. During the calculation of the action resource reallocation, a resource adjustment factor is introduced to adjust the resource allocation of each affected action parameter. The calculation of the resource adjustment factor is divided into two steps: First, divide the influence weight of the affected action parameter by the sum of the influence weights of all affected action parameters to obtain a ratio value; then multiply this ratio value by the resource remaining ratio of the underlying unit, where the resource remaining ratio is obtained by dividing the total amount of resources minus the current resource usage by the total amount of resources. The adjusted resource allocation calculation method is to multiply the original resource allocation of the affected action parameter by an adjustment coefficient, which is composed of 1 plus the resource adjustment factor, and at the same time ensure that the adjusted resource allocation does not exceed the total amount of resources. Dynamically adjust the resource allocation according to the influence weight and resource remaining situation, so that the resources are tilted towards the affected action parameters with larger influence weights, while avoiding resource allocation exceeding the total limit. The final result obtained is the adjusted resource allocation plan.

[0040] S3.4. During the calculation of the preliminary coordination plan generation, update the values of the affected action parameters according to the aforementioned adjusted resource allocation plan. The specific update method varies according to the type of the affected action parameter: for the affected action parameter of the speed type, the adjusted parameter value is obtained by multiplying the original parameter value by the ratio of the adjusted resource allocation to the original resource allocation; for the affected action parameter of the monitoring radius type, the adjusted parameter value is obtained by multiplying the original parameter value by the square root of the ratio of the adjusted resource allocation to the original resource allocation. The preliminary coordination plan includes the adjusted affected action parameter values of all underlying units. Reasonably adjust the affected action parameters through the change of resource allocation to ensure that the action capabilities of the underlying units match the change requirements of the high-level instructions. The final result obtained is a preliminary coordination plan containing all adjusted affected action parameters.

[0041] The preliminary coordination plan is used as the input for the next step to scan for cross-level conflicts and perform parameter adjustments to eliminate conflicts. Specifically, the preliminary coordination plan is passed to step S4 for detecting and resolving cross-level conflicts; the preliminary coordination plan optimized through step S4 is further used as the parameter set for step S5 to be integrated back into the object relationship graph and verify whether the expected goals of the high-level instructions are achieved through real-time simulation. By providing the set of adjusted affected action parameters, it is ensured that the subsequent steps can directly utilize the results of the current resource allocation, thus realizing the continuity of conflict resolution and plan verification. The final result obtained is that the preliminary coordination plan serves as a data bridge connecting the current step and the processing flow of the subsequent steps.

[0042] Through the above calculation logic and processing idea, step S3 is based on the set of affected action parameters in the dynamic constraint conduction chain. By calculating the influence weights, sorting them in descending order, and dynamically adjusting the action resources of the underlying units, a preliminary coordination plan is successfully generated. This plan comprehensively considers the time influence factor, space influence factor, and resource shortage degree, ensuring the rationality and efficiency of resource allocation, and providing a reliable input data basis for the cross-level conflict detection and resolution in step S4.

[0043] In step S3, a preliminary coordination plan is generated by sorting the set of affected parameters in the dynamic constraint conduction chain according to the influence weights and reallocating the action resources of the underlying units. This plan adjusts the action parameters of the underlying units, attempting to meet the spatio-temporal constraints of the high-level instructions. However, due to the hierarchical differences between the high-level instructions and the underlying units, the preliminary coordination plan may still have cross-level conflicts, that is, there are inconsistencies between the spatio-temporal coverage requirements of the high-level instructions and the action trajectories of the underlying units. The goal of step S4 is to scan the preliminary coordination plan, identify these cross-level conflicts, evaluate their severity, and adjust the action parameters of the underlying units through an iterative optimization method to resolve the conflicts, providing a conflict-free action plan for step S5.

[0044] Step S4 includes the following: S4.1, in the processing of conflict identification and spatio-temporal mapping, first map the action parameters of the underlying units in the preliminary coordination plan to the spatio-temporal coordinate system to identify the areas inconsistent with the spatio-temporal constraints of the high-level instructions. The specific calculation logic is as follows: Record the action trajectories of the underlying units within the specified time range and compare them with the coverage requirements of the high-level instructions at the same time and space positions. The criterion for identifying the conflict area is that at a certain time and space position, if the high-level instruction requires this position to be covered but no action trajectory of the underlying unit meets this requirement, then mark this position as a conflict area. Subsequently, discretize these conflict areas into spatio-temporal grids, with each grid cell corresponding to a specific time and space discrete point, and mark whether there is a conflict in the cell. Through spatio-temporal mapping and discretization, the continuous conflict areas are transformed into quantifiable grid cells for subsequent analysis of the distribution and aggregation characteristics of the conflicts. Finally, a conflict spatio-temporal grid is generated, in which the grid cells with conflicts are marked.

[0045] S4.2. When calculating the fractal aggregation index, the box-counting method is used to quantify the aggregation degree of conflicts in the spatio-temporal grid. The specific calculation logic is as follows: The spatio-temporal grid is divided into multiple boxes at different scales, and the number of boxes containing conflict grid cells is counted for each scale; then, a regression analysis is performed on the logarithmic relationship between the number of boxes at different scales and the scale, and the fractal dimension is calculated by fitting. The fractal dimension reflects the complexity and self-similarity of the conflict grid; then, the fractal dimension is divided by the embedding dimension of the spatio-temporal grid. The embedding dimension is usually the dimension number of the spatio-temporal grid. For example, the embedding dimension of a two-dimensional spatio-temporal grid is 2, so as to obtain the fractal aggregation index. The larger the value of the fractal aggregation index, the higher the aggregation degree of conflicts in space and time. The fractal theory is used to capture the irregular distribution characteristics of conflicts to more accurately evaluate the concentration degree of conflicts. The finally obtained fractal aggregation index reflects the spatio-temporal aggregation degree of conflicts.

[0046] S4.3. When calculating the conflict transfer index, first construct a conflict transfer probability matrix between underlying units to quantify the propagation trend of conflicts between underlying units. The specific calculation logic is as follows: The spatial proximity is measured by calculating the average spatial distance of the action trajectories of two underlying units. The smaller the distance, the higher the transfer probability; the similarity of action parameters is measured by calculating the difference degree of the action parameters of two underlying units. The smaller the difference degree, the higher the transfer probability; then, the transfer probability is calculated using an exponential decay function, and all transfer probabilities are normalized to ensure that the sum of the transfer probabilities of each underlying unit is 1; finally, the maximum eigenvalue of the transfer probability matrix is taken as the conflict transfer index. The maximum eigenvalue reflects the propagation potential of conflicts in the underlying unit network. The larger the value, the stronger the conflict propagation trend. By constructing the transfer probability matrix, the potential propagation paths of conflicts between underlying units are simulated, so as to evaluate the overall propagation risk of conflicts. The finally obtained conflict transfer index reflects the overall trend of conflict propagation.

[0047] S4.4. In the comprehensive conflict severity assessment, the fractal aggregation index and the conflict transfer index are combined by weighted summation to calculate the comprehensive conflict severity coefficient. The specific calculation logic is as follows: The fractal aggregation index and the conflict transfer index are linearly combined. The weights of the fractal aggregation index and the conflict transfer index are determined according to the actual scenario. For example, the weight of the fractal aggregation index is 0.6, and the weight of the conflict transfer index is 0.4. The weighted sum of the two is the comprehensive conflict severity coefficient. By comprehensively considering the aggregation characteristics and propagation trends of conflicts, the severity of conflicts is comprehensively evaluated to ensure that the evaluation results reflect both the local concentration of conflicts and their global propagation risks. The finally obtained comprehensive conflict severity coefficient is used to evaluate the overall severity of conflicts.

[0048] S4.4. When iteratively optimizing and adjusting parameters, the gradient descent method is used to optimize the comprehensive conflict severity coefficient. The goal is to adjust the action parameters of the underlying units to reduce the conflict severity. The specific calculation logic is as follows: Use the action parameters in the preliminary coordination plan as the initial values; calculate the partial derivatives of the comprehensive conflict severity coefficient with respect to each action parameter, where the partial derivative represents the direction and magnitude of the impact of parameter adjustment on the change in conflict severity; update the action parameters according to the partial derivatives and the learning rate, where the learning rate controls the step size of each adjustment; repeat this process until the comprehensive conflict severity coefficient converges or reaches the preset number of iterations; at the same time, ensure that the adjusted action parameters are within the preset feasible range to avoid exceeding the actual limits. Gradually reduce the conflict severity through iterative optimization to ensure that the adjustment of the action parameters of the underlying units can effectively resolve conflicts while maintaining the physical feasibility of the parameters. Finally, obtain the optimized set of action parameters to minimize the comprehensive conflict severity coefficient.

[0049] Use the optimized set of action parameters as the input for the next step, which is used to integrate back into the object relationship graph and verify whether the expected goals of the high-level instructions are achieved through real-time simulation. The specific calculation logic is as follows: Directly pass the optimized set of action parameters to the subsequent steps as the data basis for processing and verification. By providing a conflict-free set of action parameters, ensure that the subsequent steps can directly utilize the optimization results, satisfy the spatio-temporal constraints of the high-level instructions, and verify the effectiveness of the simulation plan through simulation. The final result is the optimized set of action parameters, which serves as the data connection basis between the current step and the subsequent steps.

[0050] Through the above processing technical logic, step S4 uses fractal aggregation analysis and conflict transfer probability assessment to accurately identify and quantify the cross-level conflict severity in the preliminary coordination plan. The action parameters of the underlying units are adjusted through the iterative optimization method, effectively resolving the conflicts and generating an optimized set of action parameters. This set ensures that the spatio-temporal constraints of the high-level instructions are satisfied, providing a conflict-free basis for integrating back into the object relationship graph and real-time simulation verification in the subsequent steps.

[0051] Step S5 includes the following: S5.1. In the process of integrating the parameter set, first map the optimized action parameter set in step S4 back to the underlying unit nodes in the multi-level object relationship graph. The specific process is as follows: for each underlying unit node in the multi-level object relationship graph, obtain its action parameter attribute, and replace the value in the preliminary coordination plan with the corresponding value in the optimized action parameter set. This replacement process ensures that the action parameters of all underlying unit nodes in the multi-level object relationship graph are completely consistent with the optimized action parameter set, thus maintaining the unity of data in the entire structure. By directly replacing the action parameter attributes of the underlying unit nodes and updating their values, the optimized action parameter set is fully integrated into the multi-level object relationship graph, providing accurate and consistent data support for subsequent simulation verification. The final output result is an updated multi-level object relationship graph, in which the action parameters of all underlying unit nodes have been adjusted to the optimized values.

[0052] S5.2. In the process of spatio-temporal coverage simulation verification, first construct a virtual spatio-temporal environment to simulate the time intervals and spatial regions of the high-level instructions. The data of these time intervals and spatial regions are derived from the spatio-temporal constraints extracted from the high-level instructions in step S1. Next, based on the action parameters of the underlying unit nodes in the updated multi-level object relationship graph, calculate the action trajectories of each underlying unit within the specified time interval. Then, conduct a point-by-point check on the spatio-temporal coverage requirements of the high-level instructions. Specifically, determine whether there is at least one action trajectory of an underlying unit that can meet the coverage requirements at each time point and spatial position. The calculation process of the coverage rate is as follows: count the number of spatio-temporal points satisfied by the action trajectories of the underlying units, and then divide it by the total number of spatio-temporal points with coverage requirements in the high-level instructions to obtain the coverage rate value. Set a coverage rate threshold, such as the standard value for full coverage. If the calculated coverage rate is greater than or equal to the threshold, it is determined that the verification passes; if the coverage rate is less than the threshold, record the uncovered spatio-temporal regions and enter the subsequent adjustment steps. Through real-time simulation and quantitative evaluation, verify whether the action trajectories of the underlying units can fully meet the spatio-temporal coverage requirements of the high-level instructions to ensure the effectiveness of the action plan. The final output results include the coverage rate value and the detailed information of the uncovered regions (if there are uncovered regions).

[0053] S5.3. In the process of local parameter fine-tuning, if the coverage rate calculated in the spatio-temporal coverage simulation verification does not reach the preset coverage rate threshold, the underlying unit action parameters corresponding to the uncovered area are locally adjusted. The specific steps are as follows: First, determine the spatio-temporal coordinate range of the uncovered area and clarify its time and space boundaries; then, screen out the underlying units that are adjacent to the uncovered area in terms of time and space; next, adjust the key action parameters of these underlying units, such as increasing the monitoring radius or raising the moving speed, to expand the coverage range of their action trajectories so that they can cover the uncovered area; after the adjustment is completed, repeat the process of spatio-temporal coverage simulation verification, recalculate the coverage rate until the coverage rate reaches the preset coverage rate threshold. By specifically modifying the action parameters of specific underlying units, gradually eliminate the uncovered area to ensure that the spatio-temporal coverage requirements of the high-level instructions are fully met. The final output result is a set of adjusted action parameters, in which the action parameters of the underlying units have been optimized to ensure that the coverage rate reaches the preset coverage rate threshold.

[0054] S5.4. In the process of generating the final action plan, integrate the set of action parameters after local parameter fine-tuning into a complete final action plan. The specific process is as follows: Based on the adjusted set of action parameters, generate the final action parameters of all underlying units and their corresponding action trajectories to ensure that these action trajectories can fully meet the spatio-temporal coverage requirements of the high-level instructions. Then, integrate and package these final action parameters and action trajectory data to form a structured collaborative execution plan, which can be directly used for subsequent simulation confrontation. By systematically organizing and packaging the optimized action parameters and their trajectory data, generate a comprehensive and executable final action plan to provide reliable support for practical applications. The final output result is a complete final action plan, which contains the final action parameters of all underlying units and the corresponding action trajectories.

[0055] Through the technical logic of the above parameter set integration, spatio-temporal coverage simulation verification, local parameter fine-tuning, and final action plan generation, step S5 realizes integrating the optimized set of action parameters back into the multi-level object relationship map, verifies the spatio-temporal coverage requirements of the high-level instructions through real-time simulation, and finally generates a final action plan that meets the expected goals. The problem of insufficient coverage is eliminated through the local parameter fine-tuning mechanism to ensure that the final action plan can fully meet the requirements of the high-level instructions, providing a solid data foundation and reliable execution basis for the collaborative execution of both sides in the simulation confrontation during the drill.

[0056] The above formulas are all dimensionless and take their numerical calculations. The formula is a formula obtained by collecting a large amount of data for software simulation to get the closest to the real situation. The preset parameters in the formula are set by technicians in this field according to the actual situation.

[0057] It should be noted that the system of the present invention can be deployed on the device itself to achieve embedded applications, or can also run on a PC or other terminals with a user interface, so as to meet various hardware environments and usage requirements.

[0058] Only some exemplary embodiments of the present invention have been described above by way of illustration. Without doubt, for those of ordinary skill in the art, various different ways can be used to modify the described embodiments without departing from the spirit and scope of the present invention. Therefore, the above drawings and descriptions are illustrative in nature and should not be construed as limiting the scope of protection of the claims of the present invention.

[0059] It should be noted that in this text, if there are relational terms such as first and second, etc., they are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variant thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements not only includes those elements, but also includes other elements not expressly listed, or also includes elements inherent to such process, method, article or device. Without further limitation, an element defined by the statement "including one..." does not exclude the existence of additional identical elements in the process, method, article or device including the element.

[0060] The above is only the specific implementation manner of the present application, but the scope of protection of the present application is not limited thereto. Any person skilled in the art within the technical scope disclosed by the present application can easily think of changes or substitutions, which should all be covered within the scope of protection of the present application. Therefore, the scope of protection of the present application should be subject to the scope of protection of the claims.

Claims

1. A method for constructing an action plan generation model based on object relationship decomposition, characterized in that: Includes steps: S1. Through object relationship decomposition technology, the spatiotemporal constraints of high-level instructions and the action parameters of low-level units are separated to construct a multi-level object relationship graph; S2. According to the multi-level object relationship graph, traverse the impact path of high-level instruction changes and extract the affected parameter set as the dynamic constraint transmission chain; S3. Sort the parameter sets in the dynamic constraint transmission chain by impact weights and redistribute the action resources of the underlying units to generate a preliminary coordination plan; S4. Scan the cross-level conflicts in the preliminary coordination plan, identify and evaluate the severity of the cross-level conflicts in the preliminary coordination plan, and use iterative optimization methods to adjust parameters to effectively resolve the conflicts; S5. Integrate the adjusted parameter set back into the object relationship graph, verify the expected goals of the high-level instructions through real-time simulation, and generate the final action plan.

2. The method for constructing an action plan generation model based on object relationship decomposition according to claim 1, characterized in that: Step S1 includes the following contents: First, the high-level instructions are parsed to extract their time constraints and space constraints, where the time constraint is defined as the time interval from the start time to the deadline of the task, and the space constraint is defined as the coordinate set of the task area; then the action parameters of each underlying unit are identified; then the high-level instructions are modeled as nodes containing time constraints and space constraints, and the underlying units are modeled as nodes containing action parameter sets, and intermediate constraint nodes are introduced to represent the sub-constraints after the high-level instruction constraint decomposition; the high-level instruction nodes are connected to the intermediate constraint nodes, and the intermediate constraint nodes are connected to the underlying unit nodes in sequence through directed edges to reflect the constraint decomposition and the restriction relationship on the underlying units; the time constraint of the high-level instructions is decomposed into multiple sub-time periods, and the space constraint is decomposed into multiple sub-areas, and the influence intensity weight of each sub-constraint on the underlying unit is calculated, where the weight calculation method is the product of the length of the sub-time period and the area of ​​the sub-area divided by the total length of the high-level instruction time interval and the total area of ​​the spatial area; finally, a multi-level object relationship graph structure containing nodes and their weighted directed edges is formed.

3. The method for constructing an action plan generation model based on object relation decomposition according to claim 2 is characterized in that: Step S2 includes the following contents: Firstly, the changes are identified by calculating the time constraint changes and space constraint changes of the high-level instructions, where the time change is defined as the sum of the offsets of the start time and the end time of the changed time interval relative to the original time interval, and the space change is defined as the sum of the non-overlapping areas of the changed spatial area and the original spatial area, so as to quantify the adjustment amplitude; then, the breadth-first search algorithm is used to traverse the impact path from the high-level instruction node through the intermediate constraint node to the bottom-level unit node, and each complete path is recorded; then, for each intermediate constraint node on the path, the time change and space change of its sub-constraints are calculated, and compared with the preset time impact threshold and space impact threshold respectively. If any change exceeds the corresponding threshold, the bottom-level unit action parameter corresponding to this path is marked as the affected parameter; finally, all affected action parameter sets are organized into dynamic constraint transmission chains, each chain includes high-level instruction nodes, affected intermediate constraint nodes and affected bottom-level unit nodes and their action parameters, so as to reflect the transmission path of the impact.

4. The method for constructing an action plan generation model based on object relation decomposition according to claim 3 is characterized in that: Step S3 includes the following contents: First, the influence weight is calculated for each affected action parameter in the dynamic constraint transmission chain. The influence weight is determined by combining the time influence factor, space influence factor and resource scarcity. The time influence factor is calculated by dividing the sub-time change of the intermediate constraint node by the length of the time interval after the high-level instruction changes. The space influence factor is calculated by dividing the sub-space change of the intermediate constraint node by the area of ​​the spatial region after the high-level instruction changes. The resource scarcity is calculated by dividing the current resource usage of the bottom unit by the total resource amount. The square root of the sum of the squares of the time influence factor and the space influence factor is multiplied by the resource scarcity adjustment term to obtain the influence weight. Then, the affected action parameter set is sorted in descending order according to the influence weight to form a priority processing list. Then, the resource adjustment factor is calculated according to the sorting result. The resource adjustment factor is obtained by multiplying the influence weight ratio of each affected action parameter by the remaining ratio of the bottom unit resources. The resource allocation is updated by multiplying the original resource allocation by the adjustment coefficient, that is, 1 plus the resource adjustment factor, while ensuring that the adjusted resource allocation does not exceed the total resource amount. Finally, the affected action parameters are updated according to the adjusted resource allocation. The specific update method is determined according to the parameter type such as speed or monitoring radius, and a preliminary coordination plan containing all adjusted affected action parameters is generated.

5. The method for constructing an action plan generation model based on object relation decomposition according to claim 4 is characterized in that: Step S4 includes the following contents: S4.1, in the process of conflict identification and spatiotemporal mapping, the action parameters of the bottom-level units in the preliminary coordination plan are first mapped into the spatiotemporal coordinate system to identify areas that are inconsistent with the spatiotemporal constraints of the high-level instructions.

6. The method for constructing an action plan generation model based on object relation decomposition according to claim 5, characterized in that: S4.2, when calculating the fractal clustering index, the box counting method is used to quantify the degree of clustering of conflicts in the space-time grid; the specific calculation logic is: divide the space-time grid into multiple boxes at different scales, and count the number of boxes containing conflict grid units at each scale; then perform regression analysis on the logarithmic relationship between the number of boxes at different scales and the scale, and calculate the fractal dimension by fitting. The fractal dimension reflects the complexity and self-similarity of the conflict grid; then, divide the fractal dimension by the embedding dimension of the space-time grid. The embedding dimension is the number of dimensions of the space-time grid, so as to obtain the fractal clustering index.

7. The method for constructing an action plan generation model based on object relation decomposition according to claim 6, characterized in that: S4.3, when calculating the conflict transfer index, firstly, the conflict transfer probability matrix between the bottom-level units is constructed to quantify the propagation trend of the conflict between the bottom-level units; The specific calculation logic is as follows: spatial proximity is measured by calculating the average spatial distance between the action trajectories of two underlying units; The action parameter similarity is measured by calculating the difference between the action parameters of two underlying units. Then the transition probability is calculated using an exponential decay function, and all transition probabilities are normalized to ensure that the sum of the transition probabilities of each underlying unit is 1. Finally, the maximum eigenvalue of the transition probability matrix is ​​taken as the conflict transition index.

8. The method for constructing an action plan generation model based on object relation decomposition according to claim 7, characterized in that: S4.4, in the comprehensive conflict severity assessment, the fractal aggregation index and the conflict transfer index are combined by weighted summation to calculate the comprehensive conflict severity coefficient.

9. The method for constructing an action plan generation model based on object relation decomposition according to claim 8, characterized in that: S4.4, when iteratively optimizing and adjusting parameters, the gradient descent method is used to optimize the comprehensive conflict severity coefficient, with the goal of adjusting the action parameters of the underlying units to reduce the severity of the conflict; the specific calculation logic is: the action parameters in the preliminary coordination plan are used as the initial values; Calculate the partial derivative of the comprehensive conflict severity coefficient with respect to each action parameter, and update the action parameters according to the partial derivative and the learning rate, where the learning rate controls the step size of each adjustment; repeat this process until the comprehensive conflict severity coefficient converges or reaches the preset number of iterations; at the same time, ensure that the adjusted action parameters are within the preset feasible range to avoid exceeding the actual limit, gradually reduce the conflict severity through iterative optimization, and finally obtain the optimized action parameter set to minimize the comprehensive conflict severity coefficient.

10. The method for constructing an action plan generation model based on object relation decomposition according to claim 9, characterized in that: Step S5 includes the following contents: First, the optimized action parameter set is mapped to the bottom unit nodes of the multi-level object relationship graph, and the action parameter attributes of each bottom unit node are updated to the corresponding optimized values ​​to maintain data consistency; Then, a virtual space-time environment is constructed to simulate and verify the space-time coverage requirements of high-level instructions in real time. Specifically, the action trajectory of the bottom-level unit is calculated by simulating the time interval and space area of ​​the high-level instruction, and it is judged whether it meets the coverage requirements of the high-level instruction at each time and space position. Then, the coverage rate is calculated, that is, the ratio of the number of space-time points satisfied by the action trajectory of the bottom-level unit to the total number of space-time points covered by the high-level instruction, and it is evaluated according to the preset coverage rate threshold. If the coverage rate does not meet the requirements, the uncovered areas are identified, the underlying units that are adjacent in time and space are screened, their key action parameters are adjusted to expand the coverage range, and the simulation verification is repeated until the coverage rate reaches the threshold; finally, the adjusted action parameter set is integrated to generate the final action parameters and trajectories of all underlying units to ensure that the time and space coverage requirements of high-level instructions are fully met and encapsulated into a final action plan that can be used to simulate confrontation.

Images