Tactical simulation deduction method based on heterogeneous computing power collaboration

Abstract

The invention discloses a tactical simulation deduction method based on heterogeneous computing power collaboration, and relates to the technical field of tactical simulation deduction, and the method comprises the following steps: constructing a simulation index structure used for distinguishing an initialized unit from a non-initialized unit, recognizing a newly-added combat unit in a simulation operation process, and carrying out the calculation of the newly-added combat unit; marking the newly added combat unit as a temporarily added combat unit based on the identification result; the method comprises the following steps: acquiring state change data and behavior instruction sequences in a plurality of continuous simulation frames on the basis of a marked temporarily-added combat unit, constructing a behavior event sequence matrix, calculating a behavior change quantity between frames by adopting a sliding window weighted summation algorithm, and establishing a short-time behavior change speed index on the behavior event sequence matrix. According to the method, the problem that the calculation power adaptation is not accurate when a combat unit is temporarily added is solved, the heterogeneous calculation power resource dynamic adjustment based on the behavior change speed is realized, and the simulation response efficiency and the time sequence precision are improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of tactical simulation deduction, and particularly relates to a tactical simulation deduction method based on heterogeneous computing power cooperation. BACKGROUND

[0002] The tactical simulation deduction based on heterogeneous computing power cooperation refers to the process of complex combat environment modeling and tactical behavior simulation, in which multiple heterogeneous computing resources (such as CPU, GPU, FPGA, etc.) are comprehensively utilized for division and cooperation to improve the real-time performance, accuracy and scalability of simulation. The existing technology of this type usually constructs a distributed simulation system architecture, first identifies the capabilities of available computing resources and manages them with labels, according to the computing density, parallelism requirements and time delay sensitivity of various simulation tasks, uses task division and scheduling algorithms to allocate high-concurrency and high-load graphics rendering, physical simulation and other sub-tasks to GPU parallel units, and processes logical control, decision-making and other serial tasks by CPU, and if necessary, uses FPGA or DSP to process low-delay and fixed-logic key modules, so as to realize cross-platform computing power cooperation. The whole simulation deduction process generally includes the following links: modeling initialization of battlefield environment and combat units, scene configuration and parameter setting of simulation tasks, scheduling and allocation of computing resources, execution and real-time interaction of multi-module parallel simulation calculation process, data synchronization and state feedback, and dynamic display and evaluation of deduction results, finally realizing rapid evolution prediction and multi-scenario deduction capability of tactical behavior through efficient cooperation of heterogeneous computing power.

[0003] The existing technology has the following deficiencies: In the process of tactical simulation deduction, when a new combat unit is temporarily added during the running of the system, the scheduler usually directly completes the initial computing resource binding according to the type information of the unit and the idle heterogeneous computing power resources in the current system. Since such a unit has not undergone actual behavior process when it is added, the system can only make a static estimation of its computing demand based on the default computing power template of the unit type, and cannot accurately reflect its real-time behavior characteristics in the subsequent simulation process. If the unit frequently triggers high-intensity tactical behavior (such as rapid state switching, path re-planning, communication cooperation, etc.) in a short time, its actual computing demand will far exceed the processing capacity of the bound resources, causing a computing bottleneck. However, the existing tactical simulation deduction technology based on heterogeneous computing power cooperation cannot dynamically judge whether the bound heterogeneous computing resources should be adjusted according to the behavior change speed of the temporarily added combat unit in the short simulation time period, so as to realize adaptive optimization of resource allocation strategy. This will cause problems such as behavior response delay and state update lag of the unit in the deduction, further affecting the timing accuracy of overall simulation synchronization and the evolution continuity of tactical behavior, and reducing the processing efficiency and simulation credibility of the system in a high-dynamic task environment.

[0004] The above information disclosed in the background section is only for the purpose of enhancing the understanding of the background of the present disclosure, and therefore it can include information that does not constitute the prior art known to those of ordinary skill in the art. SUMMARY

[0005] The purpose of the present application is to provide a tactical simulation deduction method based on heterogeneous computing power cooperation to solve the problems in the background.

[0006] In order to achieve the above-mentioned purpose, the present application provides the following technical scheme: a tactical simulation deduction method based on heterogeneous computing power cooperation, specifically comprising the following steps: S1, by constructing a simulation index structure for distinguishing initialized units from non-initialized units, identifying newly added combat units during simulation running, and based on the identification result, marking the newly added combat units as temporarily added combat units; S2, based on the marked temporary added combat units, collecting state change data and behavior instruction sequence in continuous multiple simulation frames, constructing a behavior event sequence matrix, using a sliding window weighted summation algorithm to calculate the behavior change amount between each frame, and establishing a short-time behavior change speed index on the behavior event sequence matrix; S3, inputting the short-time behavior change speed index into a resource adaptation evaluation structure, establishing a multivariate mapping relationship between the behavior change speed and the response ability of the bound heterogeneous computing resources, extracting the resource response deviation value, and generating a computing power adjustment candidate marker; S4, according to the computing power adjustment candidate marker, combining the computing power usage records and behavior change trend of the temporary added combat units in the previous several simulation frames, constructing a dynamic judgment logic path graph, and outputting a scheduling decision value of whether the heterogeneous computing power binding relationship needs to be adjusted through the conditional judgment of the graph structure; S5, according to the scheduling decision value, performing adaptive adjustment operation on the binding relationship between the temporary added combat units and the heterogeneous computing resources, including when the scheduling decision value is in the rebinding state, releasing the current binding and selecting the heterogeneous computing resources that meet the resource adaptation conditions to establish a new binding relationship.

[0007] Preferably, S1 specifically comprises: By constructing a simulation index structure for distinguishing initialized units from non-initialized units, in the simulation initialization stage, a key-value pair composed of unit identification and simulation start frame number is generated for each initialized unit, and all key-value pairs are written into the simulation index structure to form a reference index set for unit identification in the subsequent simulation running period; During the simulation running process, the unit identifier and the current simulation frame number of the combat unit entering the simulation process are extracted in real time, a comparison key-value pair is generated, and the comparison key-value pair is matched with the reference index set. When the matching result does not exist and the current simulation frame number is greater than the frame number value in any reference key-value, the newly added combat unit is identified. Based on the identification result, the unit identifier of the newly added combat unit is written into the newly added unit record set, and a temporary addition mark is added to the unit identifier, so as to perform the behavior change speed processing process on the temporarily added combat unit subsequently.

[0008] Preferably, S2 specifically comprises the following steps: S201, based on the marked temporarily added combat unit, state change data and behavior instruction sequence in multiple simulation frames are continuously collected from the simulation frame where the combat unit first appears during the simulation running process, and the collection results are sequentially composed into a time sequence set with frame numbers as the sequence, and are archived and stored according to the unit identifier for subsequent matrix construction; S202, the state change data and the behavior instruction sequence corresponding to each frame are uniformly formatted and encoded, converted into behavior event vectors of the same dimension, and all the behavior event vectors are filled into a two-dimensional structure with the unit identifier as the row index and the simulation frame number as the column index according to the time sequence, to construct a behavior event sequence matrix; S203, a sliding window structure is set on the behavior event sequence matrix, the behavior change amount between adjacent frames is calculated by using a weighted summation algorithm for the behavior event vectors in each window range, and the change amount value is time-normalized to form a short-time behavior change speed index, which is used to represent the behavior dynamic intensity of the unit in the window period.

[0009] Preferably, S203 specifically comprises: The sliding window structure is constructed on the behavior event sequence matrix at a fixed frame number interval, the width of the sliding window is equal to the number of continuous simulation frames, the window linearly slides on the behavior event sequence matrix at a frame step, and the behavior event vector set corresponding to all simulation frames in the current window range is extracted in each sliding to serve as the input basis for behavior change analysis; The vector difference operation is performed on the behavior event vectors between adjacent frames in the behavior event vector set in each sliding window, and the weighted summation algorithm is performed based on the preset behavior dimension weight configuration, so as to calculate the behavior change amount between each frame in the window range, and generate a behavior change amount sequence as an intermediate output; Divide each item in the behavior change amount sequence by the corresponding inter-frame time interval to obtain a behavior change intensity value per unit time, and perform window mean processing on the behavior change intensity value sequence to generate a short-term behavior change speed index, which is used to represent the behavior dynamic intensity change trend of the temporarily added combat unit in the current sliding window period.

[0010] Preferably, S3 specifically is: Pair the short-term behavior change speed index with the heterogeneous computing resource identifier bound to the temporarily added combat unit to construct a unit behavior-computing resource input structure, input into the resource adaptation evaluation structure, and use the unit identifier as the primary key index to ensure the independence and continuity of the unit input data in the evaluation structure; In the resource adaptation evaluation structure, call the multivariate mapping construction function to uniformly and normalize the short-term behavior change speed index and the response capability parameters of the heterogeneous computing resource in dimension, generate a multivariate mapping relationship model reflecting the correlation between the actual load of the resource and the unit behavior rate, and calculate the difference value between the current behavior change speed of the unit and the computing power response capability based on the multivariate mapping relationship model as the resource response deviation value; Compare the resource response deviation value with the preset deviation threshold interval, and when the resource response deviation value exceeds the upper limit of the preset deviation threshold interval, mark the current unit as a computing power adjustment candidate unit, and generate a computing power adjustment candidate marker containing the unit identifier and the deviation value, which is used for subsequent judgment and control operation of the resource binding relationship.

[0011] Preferably, S4 specifically includes the following steps: S401, according to the unit identifier in the computing power adjustment candidate marker, extract the computing power usage record and behavior change trend of the corresponding temporarily added combat unit in the previous several simulation frames, align the two time series according to the frame number, construct a joint feature array with the frame number as the primary index and the behavior rate and computing power occupation as the joint field, and use it as the basic input for subsequent path graph construction; S402, based on the joint feature array, construct a dynamic judgment logic path graph, use the frame number as the path node, use the relationship mode between the behavior change trend and the computing power usage record as the edge weight attribute, connect the nodes in sequence to form a graph path structure, and add a judgment condition label to each edge in the path structure to express the associated conditions of behavior surge and computing power load mismatch; S403, traverse the judgment condition labels in the dynamic judgment logic path graph, identify the path segments that continuously satisfy the edge weight condition anomaly, and perform scoring calculation according to the path length, continuous trigger times and edge weight intensity, when the score value exceeds the adjustment judgment threshold, generate a scheduling decision value, which is used to output the judgment conclusion of whether the heterogeneous computing power binding relationship needs to be adjusted.

[0012] Preferably, S403 is specifically: In the dynamic judgment logical path graph, all path nodes are traversed in time sequence, judgment condition labels on each path are extracted, and for each edge judgment condition label and edge weight parameter, edge elements meeting the abnormal relationship threshold are screened out, and their corresponding frame numbers are recorded as potential abnormal marker points; Based on the marked potential abnormal marker points, abnormal edge segments connected in the path graph are identified, the path length, abnormal condition continuous triggering times and corresponding edge weight intensity of each abnormal path segment are calculated, and a scoring function model is constructed through a weighted combination based on logistic regression to generate a path score value reflecting the resource demand pressure level; The path score value is compared with the adjusted judgment threshold, and if the path score value is higher than the adjusted judgment threshold, a scheduling decision value is generated for the temporarily added combat unit associated with the corresponding path segment, and a judgment conclusion is output that the heterogeneous computing power binding relationship of the unit needs to be adjusted, and the trigger determination of the scheduling state transition is completed.

[0013] Preferably, S5 is specifically: According to the scheduling decision value, it is judged whether the temporarily added combat unit is in the rebinding state, and when the scheduling decision value indicates the rebinding state, the temporarily added combat unit currently in the rebinding state is selected as the target unit, and the adaptive adjustment operation of the binding relationship is prepared to be executed; The binding relationship between the target unit and the currently bound heterogeneous computing resource is released, the resource identifier of the heterogeneous computing resource is deleted from the unit-resource mapping table, and the resource identifier is written into the resource idle marker set and marked as allocable state; In the resource idle marker set, the heterogeneous computing resource meeting the resource adaptation condition is searched, the adaptation degree of each candidate resource is calculated according to its response ability, historical load level and short-time behavior change speed index of the target unit, the heterogeneous computing resource with the highest adaptation degree is selected as the new binding target, the unit-resource mapping table is updated, and the adaptive adjustment operation of the binding relationship between the temporarily added combat unit and the heterogeneous computing resource is completed.

[0014] In the above technical solution, the technical effects and advantages provided by the present application are: 1. The application realizes the accurate adaptation of the computing power resources of the temporarily added combat units in the tactical simulation deduction process by constructing a heterogeneous computing power binding mechanism with dynamic scheduling capability. The scheme takes the short-time behavior change speed index as the core quantitative feature, combines the behavior event sequence matrix construction and the sliding window weighted calculation method, can extract the behavior dynamic intensity of the unit in the continuous simulation frame in real time, so as to effectively capture the behavior surge trend and overcome the problem of large error of traditional computing power static estimation. At the same time, with the help of the multivariate mapping relationship model established by the resource adaptation evaluation structure, the system can quantify the deviation between the unit behavior rate and the response ability of the currently bound heterogeneous resources, generate computing power adjustment candidate markers with controllable precision, and provide accurate basis for subsequent scheduling judgment.

[0015] 2. The application introduces a dynamic judgment logic path graph constructed based on time series, establishes an abnormal recognition mechanism between behavior change and computing power use through conditional association analysis of path nodes and edge weight attributes, and realizes intelligent evaluation of potential resource bottleneck state combined with a logistic regression scoring function. When the scoring result exceeds the adjustment judgment threshold, the system will automatically generate a scheduling decision value and trigger the adaptive adjustment operation of the resource binding relationship. This process integrates computing power use history, behavior trend dynamics, resource adaptation degree evaluation and other multidimensional factors, so that the resource binding relationship has migratory and flexible adjustment capability, significantly improves the task processing efficiency and response stability of the simulation system in a high dynamic environment, and effectively guarantees the time sequence coherence and calculation accuracy of the simulation behavior. BRIEF DESCRIPTION OF DRAWINGS

[0016] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the drawings needed in the embodiments will be briefly introduced as follows. Obviously, the drawings in the following description are only some embodiments described in the present application, and other drawings can also be obtained by those skilled in the art based on these drawings.

[0017] Figure 1 The flowchart of the tactical simulation deduction method based on heterogeneous computing power cooperation of the present application. DETAILED DESCRIPTION

[0018] Example implementations will now be described more fully with reference to the accompanying drawings. Example implementations, however, can be implemented in many different ways and should not be construed as limited to the examples set forth herein; rather, these example implementations are provided so that this disclosure will be thorough and complete, and will fully convey the inventive concept to those skilled in the art.

[0019] The present application provides a tactical simulation deduction method based on heterogeneous computing power cooperation as shown in Figure 1 The present application provides a tactical simulation deduction method based on heterogeneous computing power cooperation as shown in S1, identify the newly added combat unit in the simulation running process by constructing a simulation index structure for distinguishing initialized units from non-initialized units, and mark the newly added combat unit as a temporarily added combat unit based on the identification result; In this embodiment, S1 is specifically: In the simulation initialization stage, a key-value pair composed of a unit identifier and a simulation start frame number is generated for each initialized unit, and all key-value pairs are written into the simulation index structure to form a reference index set for unit identification in the subsequent simulation running period. In the simulation initialization stage, a unique unit identifier is assigned to each initialized unit, and the corresponding simulation start frame number when it enters the simulation system is recorded. A set of "unit identifier-frame number" key-value pairs can be constructed. This set can be implemented using a hash table, a bidirectional mapping table, or a two-dimensional array structure to support efficient data insertion and search operations. In the specific implementation, the unit configuration data at the start of the simulation can be extracted by the initialization configuration file parser to automatically generate an identifier code for each unit, such as a type prefix plus a number sequence, and then read the start frame number of the current simulation timer and bind it with the unit identifier as a key-value pair written into the data structure. Once this set is constructed, it can be used as a basis for determining the source of the unit during the simulation running process. When a new unit identifier appears in the subsequent simulation frame, the system can determine whether the unit has been registered in the initialization stage by querying this set, thereby establishing a basis for subsequent unit classification and labeling.

[0020] The simulation index structure for distinguishing initialized units from non-initialized units is a static data set constructed based on the simulation start time. Its core technical feature is to record the identity of each initialized unit and the reference time point when it first participates in the simulation operation. The function of this index structure is not only as a static reference for unit identity, but more importantly, it realizes the "generation sequence" judgment capability in the time domain. The frame number of the initialized unit usually falls in the initial segment of the simulation time axis, while the non-initialized unit, i.e., the unit added dynamically in the middle of the simulation, usually appears much later than the initialization distribution. Therefore, the timestamp comparison based on the frame number becomes a key indicator for distinguishing between the two types of units. Through the joint storage and discrimination of unit identifier and frame number, this simulation index structure can effectively support subsequent dynamic unit identification, behavior observation triggering, and computing power allocation screening, etc. high-level tasks, and is an indispensable basic component in the simulation scheduling logic.

[0021] In the simulation running process, the unit identifier and the current simulation frame number of the combat unit entering the simulation process are extracted in real time, a comparison key-value pair is generated, and the comparison key-value pair is matched with the reference index set. When the matching result does not exist and the current simulation frame number is greater than the frame number value in any reference key value, a new combat unit is identified; In the simulation running process, the unit identifier and the current simulation frame number of the combat unit entering the simulation process are extracted in real time, a comparison key-value pair is generated, and the comparison key-value pair is matched with the reference index set. When the matching result does not exist and the current simulation frame number is greater than the frame number value in any reference key value, a new combat unit is identified;

[0022] The unit identifier needs to be globally unique, which is usually generated by using a unified coding rule, for example, a combination of a task type prefix and a time stamp serial number; the simulation frame number is automatically advanced by the simulation master clock, ensuring that the time sequence of all units entering the system is comparable. The generation process of the comparison key-value pair requires that the unit identifier and the frame number be bound in a structured manner for efficient table lookup logic. The key to the matching process is the double-condition judgment mechanism, that is, not only does the unit identifier not exist in the index set, but also the current frame number must be later than the initialization distribution to accurately locate the new unit. This double-condition screening mechanism avoids the re-identification of the initialization phase units and enhances the time sensitivity and logical rigor of the identification of new units.

[0023] Based on the identification result, the unit identifier and the current simulation frame number of the unit identified as a new combat unit are written into the new unit record set, and a temporary addition mark is added to the unit identifier for subsequent processing of the behavior change speed of the temporarily added combat unit.

[0024] After the identification of the newly added combat unit is completed, the unit identifier obtained by identification and the corresponding current simulation frame number are written into a special data set for recording all units dynamically added during simulation running. The data set can be implemented by an extensible hash map structure or a linked list structure arranged in time sequence to support subsequent unit retrieval and time positioning. The writing process should ensure the uniqueness of the unit identifier and bind the current frame number as time reference information, so as to determine the starting position of the behavior change in subsequent behavior calculation. At the same time, a special logical marker is added to the unit identifier to indicate that it is a temporarily added combat unit. The marker field can be used as an input filtering condition for the behavior change speed analysis function, and only the units marked with the marker are executed for behavior tracking logic, thereby isolating the processing flow of initialization units and running period units at the data processing level, improving system running efficiency and data structure organization.

[0025] The core of this process is concentrated in three aspects: first, the dual-element binding mode of unit identifier and time information ensures that the record set has both identity indexing function and time sequence reference capability; second, the data structure selection of the record set needs to support fast search, batch update and time sorting, and the commonly used structures include time hash table with index table or ordered linked list driven by simulation frame; third, the generation and binding mechanism of the temporary addition marker, which should have unique Boolean state or level identification attribute, and be directly mounted on the unit identifier through the binding relationship, which can be used as the basis for execution branch determination in the behavior analysis stage. Through the above mechanism, the newly added unit can be effectively distinguished from the whole combat unit in logic, and accurate and efficient data basis can be provided for subsequent behavior change speed extraction and power resource adaptation calculation process.

[0026] S2, based on the temporarily added combat unit marked, collecting state change data and behavior instruction sequence in a plurality of continuous simulation frames, constructing a behavior event sequence matrix, using a sliding window weighted summation algorithm to calculate the behavior change amount between each frame, and establishing a short-time behavior change speed index on the behavior event sequence matrix; In this embodiment, S2 specifically includes the following steps: S201, based on the temporarily added combat unit marked, starting from the simulation frame where the combat unit first appears in the simulation running process, continuously collecting state change data and behavior instruction sequence in a plurality of simulation frames, and composing a time sequence set in the order of frame number, and archiving and storing by unit identifier for subsequent matrix construction; In the simulation deduction process, the unique identifier and the current frame number of the combat unit entering the simulation frame are monitored in real time. To realize the dynamic capture of the behavior of the newly added unit, the state change data and the behavior instruction sequence of the unit in multiple simulation frames are recorded continuously from the first appearance frame of the combat unit. The state change data includes dynamic attributes such as spatial position, velocity vector and attitude information, and the behavior instruction sequence includes logical information such as decision behavior type, response trigger type and action execution flag. The collection method can be based on the simulation memory callback mechanism, and the attributes and behaviors of the target unit are called through a unified data interface at each frame refresh, and a time series set is constructed according to the frame number, which is written into the corresponding data channel according to the unit identifier to form a structured archive. This processing method can ensure that the subsequent construction of the behavior event sequence matrix has a stable time axis index and consistent data dimension, which is convenient for subsequent behavior change trend analysis.

[0027] The "marked temporary added combat unit" refers to a unit that is dynamically inserted in the middle of simulation running and whose identity is confirmed through simulation index structure and matching judgment mechanism. Data of multiple simulation frames are continuously collected to construct a behavior data sequence that can reflect the evolution rule of time dimension, and it is usually recommended that the number of frames covers the complete behavior period of the unit in a short time to ensure the completeness of the characteristics. The state change data and the behavior instruction sequence are two key dimensions of the unit behavior logic and resource demand change, the former reflects the physical dynamics, and the latter reflects the logical response, and the combination of the two can restore the complete behavior trajectory of the unit in the simulation period. The collected results are sequentially composed into a time series set according to the frame number to ensure the consistency and retrievability of the data structure in the time dimension. Archiving storage according to the unit identifier provides a unique index basis for constructing the behavior event sequence matrix according to the unit dimension, thereby ensuring the accuracy and controllability of behavior change quantity extraction and speed index construction.

[0028] S202, uniformly format and encode the state change data and the behavior instruction sequence corresponding to each frame, convert them into behavior event vectors of the same dimension, and fill all the behavior event vectors into a two-dimensional structure with unit identifier as row index and simulation frame number as column index according to time sequence, to construct a behavior event sequence matrix; In order to realize the alignment of state change data and behavior instruction sequence in the time dimension in different simulation frames, the two types of data need to be uniformly formatted and encoded. Specifically, by presetting a data standard template, various parameters (such as position, speed, and attitude) in the state change data and various flags (such as instruction type, priority, and response trigger) in the behavior instruction sequence can be mapped to a fixed-dimension numerical vector, with each dimension representing a behavior characteristic item. After encoding, the behavior event vector generated by each simulation frame is sequentially stored in a two-dimensional array structure according to time order, with the row index as the unit identifier and the column index as the simulation frame number. Finally, a behavior event sequence matrix with uniform data dimensions is formed, with the unit as the main axis and the time frame as the sequence. This structure not only solves the problem of heterogeneous data alignment, but also provides a basic structural carrier for subsequent sliding window processing and behavior change calculation, significantly improving the quantifiable processing capability of behavior dynamics.

[0029] The state change data and behavior instruction sequence are derived from the continuous monitoring of the simulation engine on the combat units. The two have inconsistent dimensions and complex content, and must be structurally integrated through uniform formatted encoding. The core feature of the behavior event vector is the same dimension, with each encoded vector representing a row of behavior data in time and having consistent dimensions in space, facilitating matrix operations. In the two-dimensional structure, the unit identifier is used as the row index to achieve unit isolation processing and avoid behavior characteristic cross contamination, and the simulation frame number is used as the column index to ensure time sequence continuity and consistency, ensuring that the behavior change trend can be completely captured by the sliding window. This behavior event sequence matrix is a key bridge that abstracts time sequence behavior information from raw data into structured information, and is the core basic data structure for subsequent sliding statistics, trend identification, and computing power evaluation. The overall design reflects a high degree of attention to data computability and structural stability, providing pre-processing data support for dynamic scheduling decisions of heterogeneous computing resources.

[0030] S203, setting a sliding window structure on the behavior event sequence matrix, calculating the behavior change amount between adjacent frames in each window range using a weighted summation algorithm, and performing time normalization on the change amount value to form a short-term behavior change speed indicator, which is used to represent the behavior dynamic intensity of the unit in the window period.

[0031] A sliding window structure is set on the behavior event sequence matrix, and the behavior change amount between adjacent frames is calculated by a weighted summation algorithm in each window range, and then the change amount value is time normalized to finally form a short-time behavior change speed index. The core purpose is to accurately depict the behavior dynamic intensity change of the temporarily added combat unit in a specific time interval. Since the behavior characteristics of the newly added unit are not stable, it is difficult to reflect the real-time behavior load characteristics based on the static template directly. However, through the sliding window mechanism, continuous monitoring in the local time domain can be realized, so as to dynamically capture the behavior fluctuation rhythm of the unit in the short term. The weighted summation algorithm further enhances the sensitivity to key behavior dimensions, and highlights the influence of high-frequency or high-intensity behaviors through difference operation and weight fusion; the time normalization processing ensures the consistency of the calculation results in the time scale, effectively eliminating the interference of different simulation frame rates or sampling densities on behavior evaluation. Therefore, this processing method can quickly and objectively establish the behavior load portrait without relying on long-term historical data, providing a reasonable quantitative basis and dynamic judgment basis for subsequent heterogeneous computing resource rebinding.

[0032] In this embodiment, S203 is specifically: A sliding window structure is constructed on the behavior event sequence matrix at a fixed frame number interval, the width of the sliding window is equal to the number of continuous simulation frames, and the window slides linearly on the behavior event sequence matrix at a frame step, and in each sliding, the behavior event vector set corresponding to all simulation frames in the current window range is extracted as the input basis for behavior change analysis; The purpose of setting the sliding window structure in the behavior event sequence matrix is to capture the continuity and mutation trend of the short-time behavior change of the combat unit on the time axis. The sliding window structure divides the time interval by a fixed frame number as the width, and slides linearly at a frame step each time, ensuring frame-by-frame coverage of the behavior event vector sequence. In the implementation process, the width parameter of the sliding window can be configured and set, usually determined according to the refresh frequency of the simulation engine and the behavior response period, to ensure that each window period can cover the complete behavior characteristics of the unit in a short time. In each sliding operation, the behavior event vector set of all frames in the current window interval is extracted and used as the input for behavior change analysis. The sliding window structure is essentially a time series analysis framework that not only preserves the time continuity of the behavior data, but also provides context support for subsequent change intensity extraction through data aggregation within the window. In this way, local analysis of the behavior stability and dynamics of the unit in different time intervals can be realized, thereby realizing high-resolution behavior trend modeling and response monitoring without destroying the overall simulation structure. The key to the design of the sliding window structure lies in the coordinated setting of the frame step and the window width, which not only ensures the calculation accuracy, but also avoids redundant calculation, effectively improving the real-time performance and sensitivity of behavior fluctuation identification.

[0033] The vector difference operation is performed on the behavior event vectors between adjacent frames in the behavior event vector set within each sliding window, and a weighted summation algorithm is performed based on a preset behavior dimension weight configuration, so as to calculate the behavior change amount between each frame within the window range, and generate a behavior change amount sequence as an intermediate quantity output; In each sliding window, in order to measure the behavior change strength of the combat unit between consecutive simulation frames, a vector difference operation needs to be performed on the behavior event vector set. The specific implementation is that for each pair of adjacent simulation frames arranged in time sequence within the window, the corresponding behavior event vectors of the two frames are extracted, and the difference between the two vectors is calculated in each dimension to form a difference vector, which reflects the change amplitude of each behavior dimension of the unit between the frame pair. Then a set of preset behavior dimension weight configurations are introduced, each behavior dimension corresponds to a weight coefficient, and the weight value is set based on the importance or algorithm consumption characteristics of the behavior type, for example, the weight of the state switching dimension can be higher than that of the path point updating dimension. After the difference vector and the behavior dimension weight vector are multiplied element by element, the summation processing is performed on the result vector to obtain the weighted change total value between the frame pair, that is, the behavior change amount. In this way, the original multi-dimensional behavior change can be converted into a single numerical expression to form a behavior change amount sequence that is easy to analyze and process. Taking a specific example, if a unit has behaviors including frequent turning, speed adjustment and target locking in five frames within a window, each behavior dimension is assigned a different weight value, and the system can accurately calculate the behavior change fluctuation degree of the unit between consecutive frames through weighted difference method according to the change strength of these dimensions. This method not only considers the frequency of behavior occurrence, but also introduces the quantitative trade-off of dimension importance, so that the generated behavior change amount sequence has higher expression ability and discrimination value, providing a mathematical basis and behavior basis for generating a short-term behavior change speed indicator.

[0034] Each item in the behavior change amount sequence is divided by the corresponding inter-frame time interval to obtain the behavior change strength value within a unit time, and the behavior change strength value sequence is processed by window mean to generate a short-term behavior change speed indicator, which is used to represent the behavior dynamic strength change trend of the temporarily added combat unit within the current sliding window period.

[0035] After the extraction of the behavior change amount sequence is completed, in order to evaluate the behavior activity degree of the unit in the sliding window period, each behavior change amount needs to be divided by the corresponding inter-frame time interval, so as to obtain the behavior change intensity value in unit time, that is, the behavior fluctuation degree per second. This calculation method ensures that the behavior intensity evaluation has the time standardization characteristic, so that the behavior change amounts at different time steps have comparability. For example, in a certain sliding window, the unit generates a large behavior change amount between two frames, but the inter-frame time is long, so the change intensity value will be diluted; on the contrary, if the behavior change is concentrated in a short time, the change intensity value will be significantly improved. In order to further stabilize the intensity evaluation results and eliminate the accidental effects of local fluctuations, the unit time behavior change intensity values are combined to form an intensity sequence, and window mean value processing is performed, that is, all values in the sequence are averaged to generate a short-term behavior change speed index. This index reflects the overall dynamic activity degree of the unit in the current sliding window period, has good behavior pattern expression ability and evolution trend sensitivity. Taking the rapid completion of multiple instruction switching of a combat unit in a window period as an example, the corresponding intensity value will be continuously high, and the mean value result will also be significantly increased, so as to indicate that the unit is currently in a high behavior density stage, which is suitable as a trigger basis for resource scheduling optimization. In this way, the short-term behavior fluctuation of the unit can be described in a refined manner, and accurate and quantitative input basis is provided for subsequent resource adjustment judgment.

[0036] S3, input the short-term behavior change speed index into the resource adaptation evaluation structure, extract the resource response deviation value by establishing a multivariate mapping relationship between the behavior change speed and the response ability of the bound heterogeneous computing resources, and generate a computing power adjustment candidate marker; In this embodiment, S3 is specifically: The short-term behavior change speed index is paired with the heterogeneous computing resource identifier bound to the temporarily added combat unit to construct a unit behavior-computing power resource input structure, which is input into the resource adaptation evaluation structure, and the unit identifier is used as the primary key index to ensure the independence and continuity of the unit input data in the evaluation structure; The operation of pairing the short-time behavior change speed index with the heterogeneous computing resource identifier bound to the temporarily added combat unit can be implemented by establishing a key-value mapping table. Specifically, in the simulation deduction period, the unique identifier of each temporarily added combat unit and the bound computing resource is maintained, and by traversing the behavior change speed index queue, an input structure is formed by the behavior change speed index and the computing resource identifier. The input structure can be constructed as a multi-field data structure containing unit identifier, behavior speed value, computing resource type and current resource load state, and is input as a unit data row into the resource adaptation evaluation structure. Taking the unit identifier as the primary key index can ensure that the evaluation data corresponding to each unit is stored and updated independently in the data table, prevent cross-contamination of unit data, and improve evaluation accuracy and structure reading efficiency. The construction of this structure enables the evaluation mechanism to analyze the resource adaptation state in parallel in the unit dimension, meeting the dynamic high-concurrency evaluation requirements.

[0037] The unit behavior-computing resource input structure is a data structure for reflecting the coupling relationship between the current behavior characteristics of the combat unit and the computing resources it depends on. The core is to bind the short-time behavior change speed index with the heterogeneous computing resource identifier and the current running load state of the resource. This structure stores multi-source heterogeneous data in a unified format through standardized field definition, facilitating subsequent multi-dimensional analysis and judgment. The resource adaptation evaluation structure is a data processing structure for evaluating the matching degree of unit behavior rate and computing power response capability, usually including input receiving components, multi-variable correlation modeling modules, difference calculators and evaluation result output buffers. This structure realizes continuous quantitative evaluation of resource configuration rationality and current load adaptability by iteratively inputting behavior rate and computing power information frame by frame, and is the basic component for realizing dynamic resource judgment and rescheduling. Through the cooperative construction of the two structures, the system can realize the dynamic diagnosis capability of resource allocation state based on real-time behavior data.

[0038] In the resource adaptation evaluation structure, a multi-variable mapping construction function is called to uniformly and normalize the short-time behavior change speed index and the response capability parameters of the heterogeneous computing resource, generate a multi-variable mapping relationship model reflecting the correlation between the actual load of the resource and the behavior rate of the unit, and calculate the difference value between the current behavior change speed of the unit and the computing power response capability as the resource response deviation value based on the multi-variable mapping relationship model; The multivariate mapping construction function is called in the resource adaptation evaluation structure, and the quantitative relationship modeling between the behavior speed and the resource response capability can be realized by introducing a multi-input regression modeling or a shallow neural network. Specifically, first, the dimension uniformity processing is performed on the short-time behavior change speed index and the heterogeneous computing resource response capability parameter, and the two types of data are standardized to the same number of dimension items, so as to avoid modeling errors caused by inconsistent inputs; then, the normalized method (such as Z-score or Min-Max normalization) is used to adjust the input values from different sources to the same dimension, so as to ensure that data with different units and scales can participate in unified calculation; then, the preprocessed input vector is fed into the trained multivariate mapping model, which can be a function fitting structure trained according to a large amount of simulation historical data, and the difference value of the matching degree of the current behavior speed and the computing power response is output. The difference value is the resource response deviation value, and the larger the value is, the more the current binding resource cannot respond to the change of the unit behavior intensity in time. For example, the current behavior rate of a unit is as high as 20 times of state update per second, and the maximum processing frequency of its binding resource only supports 12 times of update per second, so the difference value is positive, and the subsequent adjustment judgment process needs to be entered.

[0039] The multivariate mapping construction function is a function mapping body for establishing the high-dimensional correlation between the unit behavior index and the computing resource response capability, which functions to jointly model multiple input dimensions and output a single dimension result, reflecting the resource matching degree of the current pairing state. The function often includes an input standardization interface, a weight fusion layer, a mapping generation engine, and an error output module. The input standardization interface is used to unify the input range and format of different data sources, and the weight fusion layer automatically assigns weight values to each input dimension according to historical training samples to ensure that important dimensions have a stronger impact on the output. The mapping generation engine is responsible for performing nonlinear combination or kernel function transformation to extract key interaction features between behavior rate and resource parameters, and finally the error output module generates the resource response deviation value and outputs it. By constructing the function, the precise evaluation of the complex behavior and resource capability relationship can be realized, supporting subsequent scheduling optimization decisions and avoiding scheduling misallocation problems caused by static rules.

[0040] The resource response deviation value is compared with the preset deviation threshold interval, and when the resource response deviation value exceeds the upper limit of the preset deviation threshold interval, the current unit is marked as a computing power adjustment candidate unit, and a computing power adjustment candidate mark containing the unit identifier and the deviation value is generated, which is used for subsequent judgment and control operation of the resource binding relationship.

[0041] The process of numerically comparing the resource response deviation value with the preset deviation threshold interval can be realized by setting upper and lower limit boundary values. The deviation value represents the response difference between the unit current behavior rate and the bound heterogeneous computing resource response capability. If the value exceeds the acceptable error range, it means that the current bound resource cannot meet the actual operation demand of the unit. In specific implementation, first, a static or dynamically generated deviation threshold interval is set, for example, a reasonable fluctuation range can be calculated based on the mean and standard deviation at the beginning of simulation. When a newly calculated resource response deviation value is generated, it is compared with the upper and lower boundaries of the interval one by one. If the deviation value is greater than the upper limit boundary value, the unit is determined to be a resource shortage risk unit, and the system will immediately add a computing power adjustment candidate marker to it. The computing power adjustment candidate marker is composed of a unit identifier and a current deviation value, which is used to identify that the unit has exceeded the carrying capacity of the current resource and needs to be processed preferentially in the subsequent scheduling process. Taking the deviation value of a unit as 0.38 and the preset upper limit as 0.30 as an example, since the deviation value is out of limit, the unit will be marked as a candidate unit and wait for subsequent computing power rebinding judgment.

[0042] The preset deviation threshold interval is a reference boundary interval for defining the acceptable error range between resource response capability and unit behavior rate. Its definition method usually includes static configuration and dynamic learning. Static configuration can be predefined during simulation platform deployment stage, for example, the experience threshold range obtained by statistical analysis of past behavior fluctuation data of the same type of unit. Dynamic learning can be adjusted online by continuously observing the historical deviation performance of the unit during simulation running, so that the interval is more suitable for real-time simulation load environment. The threshold interval is usually composed of a minimum deviation value and a maximum deviation value, and the middle part represents the normal fluctuation range between unit behavior and resource capability. When the resource response deviation value is within the interval, the system defaults that the resource matches normally and no adjustment is needed; once it exceeds the upper limit, it means that the unit may have behavior lag or response congestion, so the computing power adjustment process needs to be started. Therefore, the preset deviation threshold interval plays a core role in quantitative evaluation in computing power matching judgment, and is the basis for realizing fine resource scheduling strategy.

[0043] S4, according to the computing power adjustment candidate marker, a dynamic judgment logic path graph is constructed by combining the computing power usage records and behavior change trend of the temporarily added combat unit in the previous several simulation frames, and a scheduling decision value of whether the heterogeneous computing power binding relationship needs to be adjusted is output through conditional judgment of the graph structure; In this embodiment, S4 specifically includes the following steps: S401, adjust the unit identifier in the computing power adjustment candidate mark, extract the computing power usage record and behavior change trend of the corresponding temporarily added combat unit in the previous several simulation frames, align the two time series according to the frame number, and construct a joint feature array with the frame number as the main index and the behavior rate and the computing power occupation as the joint field, as the basic input for subsequent path graph construction; In order to realize the accurate judgment of the scheduling state, a data structure is needed to represent the coupling relationship between the dynamic behavior and the computing power resource usage state of the combat unit in a short time. The specific implementation is as follows: first, according to the unit identifier in the computing power adjustment candidate mark, the historical running data of the unit in the simulation system is retrieved, including the computing power resource index consumed by the unit in each simulation frame and the behavior change speed index. The computing power usage record can be collected by monitoring the occupation rate of the heterogeneous computing resources bound by the unit in each frame, the memory bandwidth usage, the number of thread calls and other parameters, while the behavior change trend is derived from the short-time behavior change speed index calculated in the previous stage. Then, the two time series are aligned on the unified time axis according to the frame number, so that the computing power usage value and the behavior change speed index corresponding to the same frame number are one-to-one corresponding. Taking the frame number as the main index, a set of joint feature vectors is formed for each frame, and a two-dimensional array structure is constructed as the basic input of the subsequent path graph node features. Through this array, the behavior-resource relationship of the unit in the continuous time period can be quantitatively modeled, thereby providing accurate support for the construction of edge weight and condition label in the path graph.

[0044] The computing power usage record refers to the resource condition of the heterogeneous computing resources called by the unit during operation, including but not limited to floating point operation proportion, core occupation time, cache hit rate and heterogeneous core switching times, etc. A set of indicators reflecting the load pressure of the unit. These indicators need to be collected at the frame level and recorded in real time through the resource monitoring interface. The behavior change trend refers to the change rate of the behavior intensity of the unit in a short time window. This trend is generated by combining the behavior event sequence matrix with the sliding window weighted processing, and is usually represented as the statistical slope or fluctuation range of a certain behavior vector. The joint feature array is a two-dimensional array with frame number as the main index and each item containing behavior rate and computing power occupation as two fields. This array realizes the fusion of behavior layer and resource layer data in time dimension. This structure can be used as the input baseline in graph neural network or graph traversal logic, to promote the construction of graph structure and the accuracy of edge weight assignment in dynamic judgment logic path graph, and also facilitate the time sequence analysis and control judgment of the risk of behavior surge leading to resource mismatch.

[0045] S402, construct a dynamic judgment logic path graph based on the joint feature array, in which frame numbers are used as path nodes, relationship patterns between behavior change trends and computing power usage records are used as edge weight attributes, nodes are connected in sequence to form a graph path structure, and judgment condition labels are added to each edge in the path structure to express the associated conditions of behavior surge and computing power load mismatch; The purpose of constructing the dynamic judgment logic path graph is to establish a time series driven graph structure to represent the time sequence evolution relationship between the behavior change trend of the temporarily added combat unit and the computing power usage record in the continuous simulation frame. The construction of the path graph takes the joint feature array as the input basis, and each item in the joint feature array corresponds to a simulation frame, containing the behavior change speed index and computing power resource occupation data of the frame. The frame numbers in the joint feature array are used as the node identifiers of the path graph, and the nodes corresponding to adjacent frame numbers are connected in sequence according to time to form a linear directed graph structure. On the edge generated between each two adjacent nodes, the change relationship between the behavior rate and the computing power occupation of the current frame and the previous frame is calculated, for example, the behavior rate increases and the computing power occupation does not significantly increase, such a pattern is defined as the "computing power response lag" type of edge weight. According to different relationship patterns, edge weight attributes are assigned to the edges, and judgment condition labels such as "behavior surge and resource not increased" or "behavior stable but resource suddenly increased" are added to each edge to capture the potential response mismatch state between behavior and resource. This graph structure can support subsequent anomaly pattern recognition and path scoring operations to provide structured decision-making basis for dynamic scheduling.

[0046] The dynamic judgment logic path graph is a graph structure constructed based on frame-level time series, which consists of three technical features: path nodes, edge weight attributes, and judgment condition labels. The path node is in units of simulation frame number, used to identify key state points in the simulation time series. The edge weight attribute is used to describe the coupling relationship between the behavior change trend and the computing power usage record between two adjacent nodes. The edge weight can be classified into response matching or response imbalance types by constructing a difference vector to analyze the relative relationship between behavior increment and computing power increment. The judgment condition label is a logical judgment element attached to the edge, used to identify whether a certain behavior-resource pair falls into the pre-defined abnormal judgment pattern set. These labels are generated according to the joint triggering rules of behavior change threshold and resource usage threshold, and form a set of traversable and quantifiable decision paths in the path graph. Through the dynamic judgment logic path graph, multi-dimensional modeling of the behavior-computing power dynamic relationship can be realized, providing a visual and scorable structured basis for scheduling judgment.

[0047] S403, traverse the judgment condition label in the dynamic judgment logic path graph, identify the path segment that continuously meets the abnormal edge weight condition, score according to the path length, continuous trigger frequency and edge weight intensity, and when the score value exceeds the adjustment judgment threshold, generate a scheduling decision value for outputting the judgment conclusion of whether the heterogeneous computing power binding relationship needs to be adjusted.

[0048] The judgment condition label in the dynamic judgment logic path graph is traversed, and the path segment that continuously meets the abnormal edge weight condition is identified, in order to accurately capture the relevance between insufficient computing power response and behavior surge of the combat unit in a specific time window in a high dynamic tactical simulation environment, thereby improving the pertinence and timeliness of heterogeneous computing power resource scheduling. By constructing a graph structure with frame number as node and behavior and computing power relationship as edge weight, the behavior and resource use trajectory of the unit evolving over time can be converted into an analyzable graph path pattern, and then the abnormal risk level is modeled by means of three elements of edge weight intensity, path continuity and trigger frequency in the graph path. When the score value exceeds the set adjustment judgment threshold, it means that the current behavior dynamic intensity experienced by the unit has continuously broken through the tolerance of resource response capability, which belongs to the high-risk resource matching imbalance state. At this time, the scheduling decision value is generated, which can make the system trigger the computing power rebinding operation in a logical and quantifiable manner, avoid the scheduling lag or excessive adjustment problem caused by relying on static rules or single index trigger in the traditional scheduling strategy, and fundamentally enhance the rapid adaptation ability of the system to sudden tactical demand and the running stability of the simulation task in the multi-source heterogeneous computing environment.

[0049] In this embodiment, S403 is specifically: All path nodes in the dynamic judgment logic path graph are traversed in time sequence, the judgment condition labels on each path are extracted, and for each edge judgment condition label and edge weight parameter, the edge elements that meet the abnormal relationship threshold are screened out, and their corresponding frame numbers are recorded as potential abnormal marker points; In order to identify the possible resource scheduling abnormal problems of the temporarily added combat units in the simulation process, the dynamic judgment logic path graph needs to be traversed and analyzed. In the specific implementation, the time sequence traversal algorithm can be used to visit all path nodes in the path graph in the order of simulation frame number. The judgment condition label and the edge weight parameter attached to the path edge between two adjacent nodes are extracted during each traversal. The judgment condition label is usually used to describe the logical relationship between the behavior rate and the algorithmic occupation, such as “sharp rise in behavior rate and constant resource load” or “increased resource usage but stable behavior change”. The abnormal relationship threshold is a preset logical judgment rule boundary, such as behavior change exceeding a certain value but resource response being lower than a certain proportion. These thresholds can be set according to the statistical results of historical simulation data and dynamically configured for different tactical task scenarios. In the traversal process, by comparing whether the judgment condition label of the current path edge meets the abnormal relationship threshold, the edge elements judged as abnormal relationship are screened out, and the frame number of the successor node connected by the edge is recorded as a potential abnormal marker point, which is used for subsequent scoring and scheduling judgment. In this way, the system can quickly detect the potential resource allocation lag or overload behavior and provide accurate time positioning information for scheduling decisions.

[0050] Based on the marked potential abnormal marker points, the continuously connected abnormal edge segments in the path graph are identified, the path length, the number of continuous triggering of abnormal conditions and the corresponding edge weight intensity of each abnormal path segment are calculated, and a scoring function model is constructed through a weighted combination based on logistic regression to generate a path score value reflecting the resource demand pressure level. To quantify the resource adaptation pressure that the temporarily added combat units may face in a certain period of time, it is necessary to identify continuously connected abnormal edge segments from the path graph based on potential abnormal marker points and score each abnormal path segment. In specific implementation, first, all potential abnormal marker points are time-ordered, and then the connection relationship of these points in the graph structure is checked one by one to identify those path segments that are continuous in frame number and have abnormal relationship on the edge judgment condition label. For each identified abnormal path segment, its path length (representing abnormal duration), abnormal condition continuous trigger times (representing abnormal stability) and edge weight strength (reflecting the imbalance degree of behavior and computing power) are calculated. Then, the three features are input into a weighted combination scoring model based on logistic regression for scoring. The weighted combination of logistic regression here is to set weight factors for multiple input features and combine bias terms to output normalized score values. The score value reflects the mismatch degree between the current path segment and the unit in abnormal behavior and resource response, and then represents the pressure level on the resource scheduling level. For example, if a path segment has long duration, high trigger frequency and high edge weight, its score result will be much higher than that of other segments. The reason for using a logistic regression model instead of a simple weighted average is that logistic regression can fit parameters according to the joint influence of abnormal conditions based on historical data, has stronger generalization ability and evaluation accuracy, and is suitable for various types of tactical simulation scenarios. In this way, the computing power scheduling risk can be quantitatively evaluated to provide accurate basis for subsequent scheduling decisions.

[0051] The path score value is compared with the adjustment judgment threshold value. If the path score value is higher than the adjustment judgment threshold value, a scheduling decision value is generated for the temporarily added combat unit associated with the corresponding path segment, a judgment conclusion is output that the heterogeneous computing power binding relationship of the unit needs to be adjusted, and the trigger determination of the scheduling state transition is completed.

[0052] In order to determine whether the temporarily added combat unit needs to perform the adjustment operation of the heterogeneous computing power binding relationship, the path score value needs to be compared with the adjustment judgment threshold value, so as to make a decision on the scheduling state. The implementation process includes: first, a set of adjustment judgment threshold values for distinguishing the resource carrying range and the scheduling trigger critical state are set, the threshold values can be obtained by statistical learning of the unit behavior dynamic intensity and resource response imbalance in the historical simulation data, and have a clear numerical boundary. Subsequently, the path score value generated based on the logistic regression in the previous processing step is compared with the judgment threshold value one by one, if the score value exceeds the judgment threshold value, it indicates that there is a significant response difference between the current behavior change rate of the unit and the heterogeneous computing resource bound to it, and there is a potential performance bottleneck. At this time, the system will immediately generate a scheduling decision value for the unit and record it as a unit that needs to adjust the computing power binding relationship. The scheduling decision value can be expressed in the form of state code to indicate whether to enter the rescheduling process, and is submitted to the scheduling control module to perform resource unbinding and rebinding operations. The setting of the adjustment judgment threshold value is directly related to the scheduling sensitivity and resource utilization efficiency, and the interval design needs to fully consider the tolerance upper limit between the unit behavior fluctuation amplitude and the resource adaptation ability. In large-scale high-dynamic tactical simulation, the higher threshold value strategy can effectively avoid frequent scheduling interference to the simulation stability, and in the resource shortage scene, the threshold value needs to be appropriately reduced to improve the response sensitivity. Through the comparison process, the scheduling behavior can be determined and dynamically triggered.

[0053] S5, according to the scheduling decision value, performing an adaptive adjustment operation on the binding relationship between the temporarily added combat unit and the heterogeneous computing resource, including when the scheduling decision value is a rebinding state, releasing the current binding and selecting a heterogeneous computing resource that meets the resource adaptation condition to establish a new binding relationship.

[0054] In this embodiment, S5 is specifically: According to the scheduling decision value, it is judged whether the temporarily added combat unit is in a rebinding state, and when the scheduling decision value indicates a rebinding state, a temporarily added combat unit currently in a rebinding state is selected as a target unit, and an adaptive adjustment operation of the binding relationship is prepared to be performed; In the simulation deduction system, in order to realize the dynamic management of the binding relationship between the temporarily added combat unit and the heterogeneous computing resource, a scheduling decision value parser can be preset in the scheduling module to parse the current scheduling state mark of each unit. The scheduling decision value is mapped and organized with unit identification as the key and state value as the value. The state value can include different state types such as keeping binding, rebinding, and preferential binding. When the parser judges that the state value of a certain unit is rebinding, the system extracts the unit from the current simulation running unit and adds a to-be-processed mark in the scheduling queue. This operation ensures that the system can respond to the resource adaptation mismatch problem in time, and provides a clear object range for subsequent resource unbinding and reallocation. For example, a unit is continuously evaluated as needing adjustment because its behavior rate is higher than the processing capacity of the resource, and the scheduling module identifies that its state is rebinding, so it can enter the binding adjustment process, ensuring the real-time and pertinence of resource dynamic allocation.

[0055] The scheduling decision value is a kind of dynamically generated control signal, which is derived from the judgment result of the resource adaptation evaluation structure on the change of the unit computing power demand. The temporarily added combat unit refers to the unit entity newly added in the simulation running process, which has the characteristics of non-initialization. The rebinding state is a kind of judgment condition for triggering resource unbinding and reallocation logic, which is usually output by the path graph scoring or deviation value comparison process. When the scheduling module parses the current state of the target unit as rebinding, the system sets the state mark of the unit in the running data structure and processes it preferentially in the resource scheduling process, so as to realize the orderly start of the resource reconfiguration process. This step provides a clear prerequisite for subsequent unbinding and new binding operations, and is the entry control point of the whole binding relationship adjustment process.

[0056] The binding relationship between the target unit and the currently bound heterogeneous computing resource is released, the resource identifier of the heterogeneous computing resource is deleted from the unit-resource mapping table, and the resource identifier is written into the resource idle mark set and marked as allocable state; In the simulation reasoning system, there is a unit-resource mapping relationship between the temporarily added combat units and the heterogeneous computing resources, which is usually managed through a hash mapping table or an associative array structure. To realize the adaptive adjustment operation of the binding relationship, the system needs to immediately release the binding of the target unit with the current heterogeneous computing resource after identifying that the target unit is in the re-binding state. The release operation can extract the corresponding resource identifier by taking the unit identifier as the index through the unit-resource mapping table, and perform the deletion operation of the mapping relationship to ensure that the unit is logically unbound from the original resource. Subsequently, the resource identifier that is unbound is written into a dedicated resource idle marker set, which is used to dynamically maintain the pool of all allocable resources in the idle state. In actual application, for example, when the target unit U123 is currently bound to GPU-7, and it is judged that U123 needs to be re-bound, the system immediately deletes GPU-7 from the mapping table and adds GPU-7 to the resource idle set, indicating that the resource can be matched with other units.

[0057] The target unit refers to the temporarily added combat unit currently judged by the system to need to perform the adjustment of the binding relationship, which has dynamic nature and complex behavior. The heterogeneous computing resource refers to various computing nodes that provide computing power support in the simulation system, which may include CPU, GPU, FPGA, etc. The unit-resource mapping table is a data structure used to maintain the binding relationship between the unit and the resource, which is usually stored in the form of key-value pair. The resource idle marker set is a collection type structure used by the system to manage the state of allocable resources during running. Whenever a resource is released, its identifier is added to the set and marked as re-bindable. Through this unbinding and resource marking mechanism, the system can maintain the continuity and efficiency of resource management, prevent resource leakage and redundant occupation, and ensure the accuracy and real-time performance of subsequent resource allocation.

[0058] In the resource idle marker set, the heterogeneous computing resources that meet the resource adaptation conditions are searched, and for each candidate resource, the adaptation degree is calculated according to its response ability, historical load level, and the short-time behavior change speed indicator of the target unit. The heterogeneous computing resource with the highest adaptation degree is selected as the new binding target, the unit-resource mapping table is updated, and the adaptive adjustment operation of the binding relationship between the temporarily added combat unit and the heterogeneous computing resource is completed.

[0059] In the process of dynamic scheduling of simulation tasks, in order to realize the adaptive binding adjustment between temporarily added combat units and heterogeneous computing resources, the system needs to screen and evaluate the currently available computing resources based on the resource idle mark set. The implementation is to first retrieve all unoccupied heterogeneous computing resources in the resource idle mark set, and calculate the adaptation degree of each candidate resource. The calculation of the adaptation degree takes the response capability parameters (such as execution delay, parallelism index) of the resource, the historical load level (such as the CPU or GPU utilization of the past N frames) and the short-term behavior change speed index of the target unit as input variables, constructs a weighted matching function, and outputs an adaptation score value reflecting the matching degree of unit behavior dynamics and resource response. Taking an example, the behavior change speed index of the target unit U456 in the recent sliding window is 0.87, the response capability of GPU-3 in the idle resource is high, and the historical load level is low, so the adaptation score may be 0.93, and if it is the highest among all candidates, GPU-3 is selected as the new binding target. The system then updates the binding relationship between U456 and GPU-3 in the unit-resource mapping table, realizing the closed-loop execution of the scheduling logic.

[0060] The resource idle mark set is a data set used to maintain the identification of all currently allocable heterogeneous computing resources, usually using a hash set to support fast lookup operations. The resource adaptation condition is a set of evaluation parameters for measuring whether a resource can carry the computing task of the target unit, including structure performance, response delay, current load state and other dimensions. Response capability measures the timeliness of resource processing external task requests, historical load level reflects the use of resources in the past time window, and short-term behavior change speed index is a quantitative expression of the behavior fluctuation intensity of the target unit in the recent period. The adaptation degree calculation function is a scoring mechanism based on the comprehensive evaluation of multiple input factors, usually implemented by weighted linear model, fuzzy reasoning system or machine learning model. The unit-resource mapping table is responsible for the final binding relationship record and update work, and is an indispensable core component of the resource scheduling system. Through this process, the system can maintain real-time and adaptability while realizing the computing power matching optimization of dynamic behavior units.

[0061] The above-described embodiments can be implemented in part or in whole through software, hardware, firmware or any combination thereof. When implemented in software, the above-described embodiments can be implemented in the form of a computer program product. The computer program product includes one or more computer instructions or computer programs. When loaded and executed by a computer, the computer instructions or computer programs cause the computer to perform all or part of the processes or functions described in the embodiments of the present application. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable apparatus. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another computer-readable storage medium, for example, the computer instructions can be transmitted from one website, computer, server or data center to another website, computer, server or data center through a wired or wireless (for example, infrared, wireless, microwave, etc.) manner. The computer-readable storage medium can be any available medium that can be accessed by a computer or a data storage device such as a server, data center, etc. containing one or more available medium collections. The available medium can be a magnetic medium (for example, a floppy disk, a hard disk, a magnetic tape), an optical medium (for example, a DVD), or a semiconductor medium. The semiconductor medium can be a solid-state disk.

[0062] It should be understood that the size of the sequence number of each process described above in various embodiments of the present application does not mean the order of execution, and the execution order of each process should be determined by its function and inherent logic, and should not constitute any limitation on the implementation process of the embodiments of the present application.

[0063] Those of ordinary skill in the art can realize that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be realized in electronic hardware, or a combination of computer software and electronic hardware. Whether the functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of the present application.

[0064] In several embodiments provided in the present application, it should be understood that the disclosed system and method can be implemented in other ways. For example, the above-described embodiments are only illustrative, for example, the division of units is only a logical function division, and actual implementation can have another division manner, for example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. In addition, the coupling or direct coupling or communication connection between the displayed or discussed units can be indirect coupling or communication connection through some interfaces, devices or units, and can be electrical, mechanical or other forms.

[0065] The units described as separate components may or may not be physically separate, and the components displayed as units may or may not be physical units, that is, may be located in one place, or may be distributed to multiple network units. Part or all of the units can be selected according to actual needs to achieve the purpose of the embodiment scheme.

[0066] In addition, each functional unit in each embodiment of the present application can be integrated into one processing unit, or each unit can exist physically alone, or two or more units can be integrated into one unit.

[0067] The above is only a specific embodiment of the present application, but the protection scope of the present application is not limited thereto, and any person skilled in the art can easily think of changes or replacements within the technical scope disclosed by the present application, which should be covered within the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

Claims

1. A tactical simulation and deduction method based on heterogeneous computing power collaboration, characterized in that, Specifically, the following steps are included: S1. By constructing a simulation index structure to distinguish between initialized and non-initialized units, newly added combat units are identified during the simulation process, and based on the identification results, the newly added combat units are marked as temporarily added combat units. S2. Based on the marked and temporarily added combat units, collect state change data and behavior command sequences within multiple consecutive simulation frames, construct a behavior event sequence matrix, and use a sliding window weighted summation algorithm to calculate the behavior change between each frame. Establish a short-term behavior change rate index on the behavior event sequence matrix. S3. Input the short-term behavior change rate index into the resource adaptation assessment structure, and extract the resource response deviation value by establishing a multivariate mapping relationship between the behavior change rate and the response capability of the bound heterogeneous computing resources, and generate computing power adjustment candidate tags. S4. Adjust candidate flags based on computing power, combine the computing power usage records and behavior change trends of temporarily added combat units in the first few simulation frames, construct a dynamic judgment logic path graph, and output the scheduling decision value of whether the heterogeneous computing power binding relationship needs to be adjusted through the condition judgment of the graph structure. S5. Based on the scheduling decision value, perform an adaptive adjustment operation on the binding relationship between temporarily added combat units and heterogeneous computing resources.

2. The tactical simulation and deduction method based on heterogeneous computing power collaboration according to claim 1, characterized in that, S1 specifically refers to: By constructing a simulation index structure to distinguish between initialized and uninitialized units, during the simulation initialization phase, a key-value pair consisting of a unit identifier and a simulation start frame number is generated for each initialized unit, and all key-value pairs are written into the simulation index structure to form a set of reference indexes for unit identification in subsequent simulation cycles. During the simulation process, the unit identifier of the combat unit entering the simulation process and the current simulation frame number are extracted in real time to generate a comparison key-value pair. The comparison key-value pair is then matched with the benchmark index set. When no matching result exists and the current simulation frame number is greater than the frame number value in any benchmark key value, it is identified as a newly added combat unit. Based on the identification results, the unit identifier and the current simulation frame number of the newly identified combat unit are written into the newly added unit record set, and a temporary addition mark is added to the unit identifier for subsequent processing of the behavior change rate of the temporarily added combat unit.

3. The tactical simulation and deduction method based on heterogeneous computing power collaboration according to claim 1, characterized in that, S2 specifically includes the following steps: S201. Based on the temporarily added combat units that have been marked, during the simulation operation, starting from the simulation frame in which the combat unit first appears, the state change data and behavior command sequence within multiple simulation frames are continuously collected, and the collected results are arranged into a time series set in order of frame number, and archived and stored by unit identifier for subsequent matrix construction. S202. The state change data and behavior instruction sequence corresponding to each frame are uniformly formatted and encoded into behavior event vectors of the same dimension. All behavior event vectors are then filled into a two-dimensional structure with the unit identifier as the row index and the simulation frame number as the column index according to the time sequence to construct a behavior event sequence matrix. S203. Set a sliding window structure on the behavior event sequence matrix, use a weighted summation algorithm to calculate the behavior change between adjacent frames for the behavior event vectors within each window range, and perform time normalization on the change value to form a short-term behavior change rate index, which is used to characterize the dynamic intensity of behavior within the window period.

4. The tactical simulation and deduction method based on heterogeneous computing power collaboration according to claim 3, characterized in that, S203 specifically refers to: A sliding window structure is constructed on the behavior event sequence matrix at fixed frame intervals. The width of the sliding window is equal to the number of consecutive simulation frames. The window slides linearly on the behavior event sequence matrix with a step size of one frame. In each slide, the set of behavior event vectors corresponding to all simulation frames within the current window range is extracted as the input basis for behavior change analysis. For each sliding window, a vector interpolation operation is performed on the behavior event vectors between adjacent frames in the set of behavior event vectors. A weighted summation algorithm is then performed based on the preset behavior dimension weight configuration to calculate the behavior change between each frame within the window range and generate a sequence of behavior change as an intermediate output. Divide each item in the behavior change sequence by the corresponding inter-frame time interval to obtain the behavior change intensity value per unit time. Then, perform window averaging on the behavior change intensity value sequence to generate a short-term behavior change rate index, which is used to characterize the dynamic intensity change trend of the behavior of temporarily added combat units within the current sliding window period.

5. The tactical simulation and deduction method based on heterogeneous computing power collaboration according to claim 1, characterized in that, S3 specifically refers to: The short-term behavior change rate index is paired with the heterogeneous computing resource identifiers bound to the temporarily added combat units to construct a unit behavior-computing resource input structure, which is then input into the resource adaptation evaluation structure. The unit identifier is used as the primary key index to ensure that the input data of each unit in the evaluation structure is independent and continuous. In the resource adaptation assessment structure, the multivariate mapping construction function is called to unify and normalize the short-term behavior change rate index and the response capability parameters of heterogeneous computing resources, generate a multivariate mapping relationship model to reflect the correlation between the actual load of resources and the unit behavior rate, and calculate the difference between the unit current behavior change rate and the computing power response capability as the resource response deviation value based on the multivariate mapping relationship model. The resource response deviation value is compared with the preset deviation threshold range. When the resource response deviation value exceeds the upper limit of the preset deviation threshold range, the current unit is marked as a candidate unit for computing power adjustment, and a computing power adjustment candidate tag containing the unit identifier and deviation value is generated for subsequent judgment and control of resource binding relationships.

6. The tactical simulation and deduction method based on heterogeneous computing power collaboration according to claim 1, characterized in that, S4 specifically includes the following steps: S401. Based on the unit identifier in the candidate marker for adjusting computing power, extract the computing power usage records and behavior change trends of the corresponding temporarily added combat units in the first few simulation frames, align the two time series by frame number, and construct a joint feature array with frame number as the main index and behavior rate and computing power usage as joint fields, as the basic input for subsequent path graph construction. S402. Construct a dynamic judgment logic path graph based on the joint feature array. In the graph structure, the frame number is used as the path node, and the relationship pattern between the behavior change trend and the computing power usage record is used as the edge weight attribute. The nodes are connected in sequence to form the graph path structure. In the path structure, a judgment condition label is added to each side to express the association condition between the behavior surge and the computing power load mismatch. S403. Traverse the judgment condition labels in the dynamic judgment logic path graph, identify path segments that continuously meet the abnormal edge weight conditions, calculate the score based on the path length, the number of consecutive triggers and the edge weight strength, and generate a scheduling decision value when the score value exceeds the adjustment judgment threshold. This value is used to output the judgment conclusion on whether the heterogeneous computing power binding relationship needs to be adjusted.

7. The tactical simulation and deduction method based on heterogeneous computing power collaboration according to claim 6, characterized in that, S403 specifically refers to: In the dynamic judgment logic path graph, all path nodes are traversed in chronological order, the judgment condition labels on each path are extracted, and for each edge's judgment condition label and edge weight parameter, edge elements that meet the abnormal relationship threshold are selected and their corresponding frame numbers are recorded as potential abnormal marker points. Based on the marked potential anomaly markers, identify the continuously connected abnormal edge segments in the path graph, calculate the path length, number of consecutive triggers of the abnormal condition, and corresponding edge weight strength of each abnormal path segment, and construct a scoring function model by weighted combination based on logistic regression to generate a path score value that reflects the level of resource demand pressure. The path score is compared with the adjustment judgment threshold. If the path score is higher than the adjustment judgment threshold, a scheduling decision value is generated for the temporarily added combat unit associated with the corresponding path segment. The judgment conclusion that the heterogeneous computing power binding relationship of the unit needs to be adjusted is output, and the scheduling state transition is triggered.

8. The tactical simulation and deduction method based on heterogeneous computing power collaboration according to claim 1, characterized in that, S5 specifically refers to: The scheduling decision value determines whether the temporarily added combat unit is in a rebinding state. When the scheduling decision value indicates that it is in a rebinding state, the temporarily added combat unit that is currently in a rebinding state is selected as the target unit, and the adaptive adjustment operation of the binding relationship is prepared to be executed. Remove the binding relationship between the target unit and the currently bound heterogeneous computing resources, delete the resource identifier of the heterogeneous computing resources from the unit-resource mapping table, and write the resource identifier into the resource idle tag set to mark it as allocable; Search the resource idle marker set for heterogeneous computing resources that meet the resource adaptation conditions. For each candidate resource, calculate the adaptation degree based on its response capability, historical load level and the short-term behavior change rate index of the target unit. Select the heterogeneous computing resource with the highest adaptation degree as the new binding target, update the unit-resource mapping table, and complete the adaptive adjustment operation of the binding relationship between the temporarily added combat unit and the heterogeneous computing resource.

Images