A game simulation performance dynamic evaluation method and system based on an entropy value method

Abstract

This invention discloses a dynamic evaluation method and system for game simulation performance based on entropy value method, belonging to the field of intelligent collaborative control technology. The method includes: constructing a prior knowledge base containing typical task sets, an evaluation index system, and performance functions; in simulation, identifying formation tasks in real time by parsing dynamic instructions and node states, and matching corresponding index systems for targeted data collection; dynamically calculating the aggregation weights of each level of indexes based on the real-time collected data using the entropy value method; and finally, using dynamic weights to weighted aggregate the index performance values ​​and outputting real-time performance evaluation results. This invention overcomes the problem of poor adaptability of traditional static evaluation methods in dynamic game scenarios, achieving dual dynamic adaptation of the index system and weights, significantly improving the real-time performance, objectivity, and accuracy of performance evaluation in large-scale autonomous simulation.

Description

Technical Field

[0001] This invention relates to the field of intelligent collaborative control technology, and in particular to a dynamic evaluation method and system for game simulation performance based on entropy method. Background Technology

[0002] Game theory simulation optimizes decision-making and resource allocation by simulating, analyzing, and evaluating the dynamic behavior of multiple parties in complex systems. It treats drones, personnel, vehicles, materials, information, and funds as intelligent agents or entities with attributes and behaviors, simulating their dynamic interactions and state changes over time in a virtual space with defined rules and constraints (such as drone delivery, urban road networks, markets, and power grids). This generates a large amount of data, allowing for the evaluation of the consequences of different strategies, identifying potential problems before implementation, optimizing solutions, reducing risks, and improving efficiency. For example, in urban management, simulations can model the impact of weekday morning rush hour traffic flow and emergencies (such as accidents and weather) on the road network, dynamically adjusting traffic light timings, predicting congestion points, planning detour routes, and evaluating the effectiveness of new roads or traffic policies (such as traffic restrictions). In drone logistics or rescue operations, real challenges arise from weather, airspace, and sudden order changes. In this scenario, a fleet of logistics drones acts as the "red team," while severe weather, sudden traffic control, and urgent orders act as interference from the "blue team," using game theory to help the dispatch center make better drone logistics or rescue decisions.

[0003] In game simulation performance evaluation, traditional performance evaluation paradigms are generally based on pre-defined scenarios, processes, and rules. This paradigm first requires pre-determining detailed game scenarios and designing standardized simulation processes accordingly. It then maps system capabilities and task performance through specific game activities, thereby constructing a static evaluation index system. During simulation, collected data is associated with final-level indicators according to pre-defined mapping relationships, and fixed aggregation weights and calculation methods are used to ultimately generate top-level quantitative results of game capability and task performance evaluation results.

[0004] However, as simulation technology evolves towards practical, large-scale, and highly dynamic game scenarios (such as autonomous simulation based on multi-agent systems), the aforementioned traditional performance evaluation paradigm has revealed a series of significant shortcomings: 1. Rigid pre-set processes and insufficient scenario adaptability: Large-scale autonomous simulation emphasizes the autonomous interaction and dynamic emergence of situations by intelligent agents under rule constraints. The game process is difficult to pre-set completely and accurately before simulation. This makes it difficult, or even impossible, to design game activities and related evaluation indicators based on pre-set processes.

[0005] 2. Limitations of static evaluation index system: Since the dynamic and ever-changing game process cannot be fixed in advance, the static evaluation index system that relies on a fixed process is difficult to effectively capture, measure and reflect the new game behaviors that emerge dynamically during the simulation process and their actual effectiveness.

[0006] 3. Lack of adaptability of the evaluation model to uncertainty: In large-scale autonomous simulation, the autonomous decision-making, interaction results, and overall game situation of the agents are all highly uncertain. Using a pre-set, fixed evaluation model (including index aggregation weights and calculation methods) for result aggregation calculation is difficult to effectively reflect and adapt to the complexity of this dynamic evolution, resulting in serious challenges to the robustness, reliability, and real-time performance of the evaluation results.

[0007] Therefore, there is an urgent need for a real-time performance evaluation technology that can dynamically and adaptively adjust the evaluation index system and aggregate weights for large-scale, highly dynamic autonomous simulation scenarios. Summary of the Invention

[0008] The purpose of this invention is to overcome the problems existing in the prior art and to provide a dynamic evaluation method and system for game simulation effectiveness based on the entropy method.

[0009] The objective of this invention is achieved through the following technical solution: Firstly, this paper provides a dynamic evaluation method for game simulation effectiveness based on entropy value method, such as... Figure 1 As shown, it includes the following steps: S1. Based on historical simulation data, establish a set of typical tasks and actions in the game scenario through cluster analysis; S2. For each type of typical task, establish a multi-level evaluation index system and set an effectiveness function for the final level index; S3. During the simulation, the status information and dynamic instructions of the simulation nodes are acquired in real time; based on the dynamic instructions and / or node status information, the typical tasks and action sets are matched to identify the task type currently being executed by each formation.

[0010] S4. Based on the identified task type, load the corresponding multi-level evaluation index system subset for each formation, and collect data from relevant nodes in a targeted manner based on the multi-level evaluation index system subset; S5. Using the performance function, calculate the performance value of mapping the data collected in step S4 to the final-level indicator; S6. Using the efficiency values ​​of the final-level indicators as samples, the efficiency of each formation in performing the current task is dynamically calculated using the entropy method.

[0011] Specifically, during the simulation, both the red and blue teams make autonomous decisions and take actions based on the current simulation situation information. By reading the task switching instructions or certain key signals within and between the red team formations, they identify the tasks currently being performed by each red team formation. By comparing these tasks with the set of typical tasks and actions established in the early stages, they determine which set of previously established indicator systems is currently suitable, and map the relevant data within the current formation to the corresponding final-level indicators and record the data.

[0012] In large-scale simulated game scenarios, multiple sub-teams are typically formed. Dynamic commands and signals enable task identification and evaluation index matching for individual teams. However, calculating the system's performance requires weight allocation for each sub-task based on the game scenario. To achieve dynamic allocation of aggregated weights in game scenarios, this invention proposes to dynamically determine the aggregated weights of each index using the entropy method. Based on the weights of each sub-task, the current task performance evaluation result can be calculated in real time.

[0013] The entropy method can be broken down into multiple steps, facilitating incremental computational optimization and adapting to data stream processing. Furthermore, this method is independent of the absolute number of samples, providing meaningful weight allocation even with small sample sizes.

[0014] In some embodiments, the steps further include: S7. Calculate the information entropy of each formation when performing the same type of task using the entropy method, and then sum the task performance of each formation based on the information entropy to obtain the comprehensive performance of performing the task.

[0015] In some embodiments, establishing a set of typical tasks and actions in a game scenario through cluster analysis includes: Extract numerical feature data that characterizes the task's properties from historical simulation data; The numerical feature data is normalized. The number of clusters is determined by considering the complexity of the numerical feature data and the actual needs. The clustering algorithm is executed to divide the action sequence space into a finite number of clusters, each cluster representing a typical task pattern.

[0016] In some embodiments, setting a performance function for the final-level indicator includes: Clearly define whether each final-level indicator is a positive or negative indicator; Determine the value range for each final-level indicator; Based on the changing patterns of the final-level indicator data and its relationship with efficiency, select an appropriate function type.

[0017] In some embodiments, the step of dynamically calculating the performance of each formation in performing the current task using the entropy method includes: The weights of the three-level evaluation indicators were calculated using the entropy method. The entropy method is used to aggregate the weights of all tertiary evaluation indicators into the weights of secondary evaluation indicators; The entropy method is used to aggregate the weights of all secondary evaluation indicators into the weights of the primary evaluation indicators; The effectiveness of the formation in performing the current task is calculated based on the weights of the primary evaluation indicators.

[0018] In some embodiments, the entropy method includes: Calculate the weight of each indicator in each sample; The information entropy of a single indicator is calculated based on the aforementioned proportion. The weight of each indicator is calculated based on information entropy; Specifically, a sliding window mechanism is used to update the sample set used for calculation, so as to achieve dynamic updating of weights.

[0019] Secondly, a dynamic evaluation system for game simulation effectiveness based on the entropy method is provided, including: The typical task and action set construction module is used to build typical task and action sets in game scenarios based on historical simulation data and through cluster analysis. The performance function design module is used to establish a multi-level evaluation index system for each type of typical task and to set performance functions for the final level index. The node information acquisition and task type identification module is used to acquire the status information and dynamic instructions of simulation nodes in real time during the simulation operation; based on the dynamic instructions and / or node status information, it matches the typical tasks and action sets to identify the task type currently being executed by each formation. The data-oriented acquisition module is used to load a corresponding multi-level evaluation index system subset for each formation according to the identified task type, and to collect data from relevant nodes in a targeted manner based on the multi-level evaluation index system subset; The final-level indicator performance value calculation module is used to calculate the performance value of mapping the data collected in the data-oriented acquisition module to the final-level indicator using the performance function. The formation performance calculation module is used to dynamically calculate the performance of each formation in performing the current task using the performance value of the final-level index as a sample and the entropy method.

[0020] It should be further noted that the technical features corresponding to the above-mentioned options and embodiments can be combined or substituted with each other to form new technical solutions without conflict.

[0021] Compared with the prior art, the beneficial effects of the present invention are: This invention pre-constructs a multi-level evaluation index system and a set of typical tasks and actions. Combined with dynamic instructions between formations or nodes during large-scale simulations, it identifies and matches the current formation's task state. From a limited task set, it selects corresponding or most suitable sub-tasks and index systems, collects simulation data in a targeted manner, generates final-level index calculation results, and then allocates the weights of each sub-task aggregation to the overall performance based on the entropy values ​​of each formation's evaluation data. This method, based on dynamic task instruction recognition and the entropy method, enables dynamic switching and autonomous allocation of the index system and aggregation weights. It addresses the objective difficulty of static evaluation methods being unsuitable in large-scale dynamic game environments, providing a paradigm for automated dynamic evaluation targeting two key aspects: index system construction and index aggregation weights. This lays the foundation for future intelligent and adaptive closed-loop optimization of game performance.

[0022] 1. Achieved dynamic self-adaptation of the evaluation system: Through the "dynamic instruction recognition - typical task matching" mechanism, the problem that the static indicator system cannot adapt to the dynamic emergent game process is solved, and the most suitable subset of evaluation indicators can be flexibly switched and matched according to the real-time situation of the simulation.

[0023] 2. Improved objectivity and real-time performance of weight allocation: The entropy method is innovatively introduced into the real-time evaluation process. The weight of the indicators is dynamically calculated by utilizing the information content of the evaluation data itself, avoiding the bias of subjectively preset weights. Moreover, this method supports incremental calculation, can adapt to small samples and realize real-time updates of weights, so that the evaluation results can better reflect the importance distribution of the current game situation.

[0024] 3. Improved automation and intelligence of the evaluation process: This invention organically integrates the construction of indicator system, task identification, data collection, weight calculation, and performance aggregation, forming a complete automated real-time evaluation paradigm. This reduces the reliance on manual intervention and lays the technical foundation for future intelligent and adaptive game performance closed-loop optimization.

[0025] 4. Enhanced evaluation capabilities in complex scenarios: By introducing formation coordination instruction recognition and multi-formation efficiency fusion steps, it can effectively handle complex simulation scenarios of large-scale, multi-formation cooperative games, making the evaluation results more comprehensive and accurate. Attached Figure Description

[0026] Figure 1 This is a flowchart of a dynamic evaluation method for game simulation effectiveness based on entropy value method according to the present invention. Figure 2 This is a detailed flowchart of an embodiment of the method of the present invention. Detailed Implementation

[0027] The technical solution of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. The components of the embodiments of this application described and shown in the accompanying drawings can generally be arranged and designed in various different configurations. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0028] It should be noted that the defects in the solutions in the prior art are all the results of the inventors' practice and careful research. Therefore, the discovery process of the above problems and the solutions proposed by the embodiments of this application in the following text should be the inventors' contributions to this application in the process of invention and creation, and should not be understood as technical content known to those skilled in the art.

[0029] In view of the technical problems pointed out in the background art, the present invention provides the following embodiments: In one exemplary embodiment, a dynamic evaluation method for game simulation performance based on entropy value method is provided. This method forms a process of "indicator system and task set construction - dynamic command task matching in formation - dynamic weighting using entropy value method," realizing performance evaluation calculation for autonomous game inference simulation in large-scale scenarios. Figure 2 As shown, it specifically includes: Step 1: Establish a priori database by inputting the simulation data corresponding to the system into the historical simulation database. The state of a single node is finite and pre-defined. Specifically, this includes collecting action trajectory data, equipment status parameters, communication command records, and game environment data of both the red and blue teams in past simulations. The collected data undergoes spatiotemporal alignment processing to unify timestamps and coordinate systems, ensuring data consistency. To identify task types through data, the K-Means clustering algorithm is used to perform cluster analysis on the preprocessed behavioral sequence data, dividing the behavioral sequence space into a finite number of typical task patterns. The specific steps are as follows: (1) Feature selection and construction: Extract numerical features that characterize the task characteristics from the original simulation data (such as the average speed and displacement of a unit within a certain time period). (2) Normalize the numerical feature data using the Min-Max method, scaling the minimum to maximum values ​​of the data to the range of [0, 1] to ensure that all data are standardized. (3) Determine the number of clusters k based on the data complexity and actual needs. (4) Perform K-Means clustering, randomly select K center points, calculate the distance from each data point to the K cluster centers (usually using Euclidean distance), and assign it to the nearest cluster. Recalculate the mean of each cluster and use the mean as the new cluster center. Repeat this step until the cluster centers no longer change significantly or the preset maximum number of iterations is reached. The cluster analysis can then be considered complete.

[0030] At this point, each data cluster can be considered to correspond to a typical task. This process can determine the typical task of a single formation by using the state information of all nodes within the formation.

[0031] Step 2: Based on established typical tasks, offline manual analysis of task processes and game-theoretic activities is conducted. For each type of typical task, its action sequence (e.g., reconnaissance-decision-attack-assessment) is decomposed to identify key performance influencing factors. A three-level evaluation index system is established for different types of tasks to evaluate task execution. The first-level index is a definite value ranging from [0,1], which typically requires comprehensive consideration of task completion rate, task completion time, and task cost-effectiveness. Due to the need to consider multiple factors, it is difficult to quantify it using a single data point at a single moment. To obtain the top-level first-level index, it needs to be decomposed layer by layer into measurable, manageable, and actionable specific indicators. Similarly, after decomposing the first-level index into second-level indicators, the second-level indicators need to be further decomposed into quantifiable third-level indicators. The third-level indicators usually directly correspond to a specific value during the task process and can be directly collected from the simulation system or equipment, or the result of simple processing of several values.

[0032] Step 3: After obtaining the three-level indicator system, in order to unify the indicator data of different dimensions and ranges into a comparable and standardized "performance score," based on this indicator system and typical task scenarios, the upper and lower bounds of the original data and the function type corresponding to each final-level indicator are determined through the analysis of sub-tasks and formation actions in typical task scenarios. The specific steps are as follows: (1) Clarify whether each indicator is a positive or negative indicator. (2) Based on prior knowledge or typical task scenario analysis, determine the reasonable value range of each final indicator in reality. Set its lower limit and upper limit. (3) Based on the changing patterns of indicator data and its relationship with performance, select an appropriate function type for calculation. Finally, use the performance function to evaluate each data in real time, such as final indicators like flight speed, flight altitude, and response time.

[0033] Step 4: Develop supporting data collection standards, and clarify the data source, collection frequency and preprocessing requirements for each indicator.

[0034] Step 5: Conduct a game simulation experiment (compatible with purely virtual or virtual-real hybrid approaches) to obtain the status and communication information of each red team node in real time. To accurately obtain the formation and grouping of unmanned nodes, it is necessary to distinguish the formation of each node through real-time dynamic commands.

[0035] Step 6: Furthermore, to address the rapid task switching during real-time evaluation, real-time dynamic commands have higher weight when identifying formation tasks. Upon receiving a clear task switching command, such as when the formation is currently in the search phase, the formation can be identified as having switched to another typical task after most nodes receive the attack command, without needing to undergo a period of data acquisition and analysis before classification. This allows for a more accurate match between task phases and data sources.

[0036] In addition, in multi-formation scenarios, a task state machine is maintained for each independent formation. When a coordination command between formations is detected (such as "formation A supports formation B"), a coordination task identification process is initiated: first, the task type for each formation is identified individually, and then the overall task mode is determined through the coordination relationship graph (a pre-established formation collaboration rule library). The identification results are matched with the indicator system library, and a corresponding subset of evaluation indicators is loaded for each formation.

[0037] Step 7: Using the defined data collection list, collect data from each node and store the data in groups. This includes fields such as timestamp, group ID, and metric values.

[0038] Step 8: Map the collected data to the final-level indicators. Calculate the performance value of mapping the collected data to the final-level indicators using a pre-established performance function.

[0039] Step 9: Based on the pre-established indicator system, to obtain the performance value of each formation's current task, all tertiary indicators need to be weighted to obtain the secondary indicator results. Similarly, the secondary indicators are weighted and summed to obtain the primary indicator results. Traditional methods use pre-set weights or allocate weights based on task execution after all tasks are completed, which is difficult to meet real-time evaluation requirements. Therefore, this patent uses the entropy method to calculate the aggregate weights of multi-level indicators during task execution, thus enabling the real-time acquisition of indicator aggregate weights after partial data generation, ultimately achieving real-time evaluation. The detailed process of calculating indicator aggregate weights using the entropy method is as follows: First, the results are calculated using a three-level index. These results are the values ​​obtained by mapping the original data through a performance function, defined as Z, which is a value located in the interval [0,1]. i Let be the i-th third-level indicator. Then calculate the weight P of this indicator. i =Z i / sum(Z). Next, calculate the information entropy E. i =-k*sum(P i *ln(P i ), where k = 1 / ln(n), and n is the number of samples. Finally, the weights W are obtained. j =(1-E i ) / sum(1-E i To adapt to dynamic changes, a sliding window update mechanism (window size = 100) is adopted. Each time new data is added, the oldest data point is discarded, and the weights within the window are recalculated. Similarly, after obtaining the aggregated weights of the tertiary indicators, the results of all secondary indicators can be obtained. The entropy method is repeatedly used to obtain the weights of the aggregation from secondary to primary indicators in the current state. This method can be used to obtain the aggregated weights from tertiary to secondary and from secondary to primary indicators.

[0040] Step 10: Performance-weighted aggregation calculation. Implementation method: For each formation, calculate according to formula S. k =sum(W i *Z i ) Calculate the subtask performance, where W i Z represents the aggregate weight corresponding to the i-th third-level indicator. i Let be the performance value of the i-th indicator. Based on the weights obtained in Step 9, first calculate the weights of all tertiary indicators aggregated into secondary indicators to obtain the performance value of the secondary indicators. Then, based on the weights of the secondary indicators aggregated into primary indicators, finally obtain the performance value of the formation in performing this task.

[0041] Step 11: After completing the above steps, the task performance of each formation can be calculated after the simulation task is completed. To address the issue of task performance calculation when multiple formations are performing the same type of task in the same scenario, the entropy of all information in each formation under that task can be obtained using the entropy method. The specific calculation steps are the same as in Step 9.

[0042] Information entropy can represent the activity level of a formation to a certain extent and can objectively reflect the formation's contribution to the task. Finally, by weighting and summing the task performance values ​​of each formation based on the information entropy, the overall performance value of the task in the current scenario can be obtained.

[0043] In another exemplary embodiment, a dynamic evaluation system for game simulation performance based on the entropy method is provided, comprising: The typical task and action set construction module is used to build typical task and action sets in game scenarios based on historical simulation data and through cluster analysis. The performance function design module is used to establish a multi-level evaluation index system for each type of typical task and to set performance functions for the final level index. The node information acquisition and task type identification module is used to acquire the status information and dynamic instructions of simulation nodes in real time during the simulation operation; based on the dynamic instructions and / or node status information, it matches the typical tasks and action sets to identify the task type currently being executed by each formation. The data-oriented acquisition module is used to load a corresponding multi-level evaluation index system subset for each formation according to the identified task type, and to collect data from relevant nodes in a targeted manner based on the multi-level evaluation index system subset; The final-level indicator performance value calculation module is used to calculate the performance value of mapping the data collected in the data-oriented acquisition module to the final-level indicator using the performance function. The formation performance calculation module is used to dynamically calculate the performance of each formation in performing the current task using the performance value of the final-level index as a sample and the entropy method.

[0044] Based on the above methods and systems, a dynamic evaluation method for the simulation and evaluation of the specific performance of large-scale urban emergency medical supply drone delivery is presented.

[0045] 1. Role Mapping: The red team refers to the intelligent drone delivery fleet and its dispatch and command system; the blue team refers to the dynamically changing civilian environment complex, including: Meteorological environment: sudden crosswinds, rainfall, low visibility (“natural interference”). Airspace management and traffic: temporary no-fly zones, manned aircraft activity, high-rise building clusters (“rules and physical interference”). Sudden changes in mission requirements: addition of emergency delivery points, surge in demand at a certain point, order cancellations (“target dynamic interference”). System inherent risks: sudden drone malfunctions, unstable communication links (“endogenous interference”).

[0046] 2. Simulation and Real-time Performance Evaluation Process Step 1: Constructing a prior knowledge base and indicator system (preparation before the game) Construction of typical task sets: Through historical data clustering analysis, typical civilian task modes such as "grid-based routine inspection", "multi-point cyclical supply", "single-point rapid delivery" and "multi-team collaborative assault" are defined.

[0047] Indicator System Establishment: A three-tiered indicator system was established for the "rapid delivery" task. For example: Primary indicator: Overall efficiency of medical supplies distribution.

[0048] Secondary indicators: timeliness, reliability, and security.

[0049] Level 3 indicators: flight speed deviation, route compliance rate, remaining battery range safety margin, and response time to abnormal weather.

[0050] Step Two: Real-time Task Identification and Dynamic Matching (Situational Awareness in Game Theory) The command center issued an order to Formation A: "Deliver vaccines to Community Hospital G, highest priority." (Explicit instruction from the Red Team) Meanwhile, the meteorological system issued a warning: "Strong gusts are expected in the 5th airspace." (Blue team interference command) The evaluation system captures these two commands in real time. It identifies that Formation A's mission has switched from "cruising" to "high-priority, disturbance-resistant delivery". The system immediately retrieves the evaluation subset that best matches this mission from the indicator library, in which the preset weight values ​​of indicators such as "wind resistance" and "route dynamic adjustment agility" are activated.

[0051] Step 3: Dynamic weight calculation based on entropy method (core of intelligent decision-making) During mission execution, the system continuously collects data. Assume that initially, "flight speed" and "course accuracy" are given high weight.

[0052] When the drone swarm entered a strong gust zone (where the blue team's interference took effect), the data stream showed a sharp divergence in the "attitude stability rate" index of each drone (some drones were greatly affected, while others were less so). According to the entropy method, the data dispersion of this index suddenly increased (entropy increase). The system automatically calculated and significantly increased the aggregation weight of "attitude stability rate" in the current evaluation, because it became a key variable that distinguished the effectiveness of individual drones within the formation and affected the success or failure of the overall mission.

[0053] Conversely, in calm airspace, the "attitude stability rate" data is highly consistent (low entropy), and its weight automatically decreases, giving way to indicators such as "flight speed".

[0054] Step 4: Real-time performance aggregation and decision support (game theory evaluation and command) The system calculates the dynamic performance value of formation A's current task in real time and presents it in the form of a dashboard. When the performance value decreases due to strong winds, the system can trigger an alarm.

[0055] More advanced applications: The command center issues a new instruction: "Formation B proceeds to point X to pick up the cargo and jointly complete the delivery." The system identifies the "multi-formation collaboration" sub-task, activates the collaboration evaluation model, and may dynamically allocate their contribution weights in the overall task based on the real-time performance data entropy of formations A and B, thereby providing a fused global performance evaluation.

[0056] The above detailed embodiments are a description of the present invention. It should not be considered that the specific embodiments of the present invention are limited to these descriptions. For those skilled in the art, several simple deductions and substitutions can be made without departing from the concept of the present invention, and all of these should be considered to fall within the protection scope of the present invention.

Claims

1. A dynamic evaluation method for game simulation effectiveness based on entropy value method, characterized in that, Includes the following steps: S1. Based on historical simulation data, establish a set of typical tasks and actions in the game scenario through cluster analysis; S2. For each type of typical task, establish a multi-level evaluation index system and set an effectiveness function for the final level index; S3. During the simulation, the status information and dynamic instructions of the simulation nodes are acquired in real time; based on the dynamic instructions and / or node status information, the typical tasks and action sets are matched to identify the task type currently being executed by each formation; S4. Based on the identified task type, load the corresponding multi-level evaluation index system subset for each formation, and collect data from relevant nodes in a targeted manner based on the multi-level evaluation index system subset; S5. Using the performance function, calculate the performance value of mapping the data collected in step S4 to the final-level indicator; S6. Using the efficiency values ​​of the final-level indicators as samples, the efficiency of each formation in performing the current task is dynamically calculated using the entropy method.

2. The method for dynamic evaluation of game simulation effectiveness based on entropy method according to claim 1, characterized in that, It also includes the following steps: S7. Calculate the information entropy of each formation when performing the same type of task using the entropy method, and then sum the task performance of each formation based on the information entropy to obtain the comprehensive performance of performing the task.

3. The method for dynamic evaluation of game simulation effectiveness based on entropy value method according to claim 1, characterized in that, The establishment of typical task and action sets in a game-theoretic scenario through cluster analysis includes: Extract numerical feature data that characterizes the task's properties from historical simulation data; The numerical feature data is normalized. The number of clusters is determined by considering the complexity of the numerical feature data and the actual needs. The clustering algorithm is executed to divide the action sequence space into a finite number of clusters, each cluster representing a typical task pattern.

4. The method for dynamic evaluation of game simulation effectiveness based on entropy method according to claim 1, characterized in that, The process of setting performance functions for final-level indicators includes: Clearly define whether each final-level indicator is a positive or negative indicator; Determine the value range for each final-level indicator; Based on the changing patterns of the final-level indicator data and its relationship with efficiency, select an appropriate function type.

5. The method for dynamic evaluation of game simulation effectiveness based on entropy value method according to claim 1, characterized in that, The method of dynamically calculating the performance of each formation in performing the current task using the entropy method includes: The weights of the three-level evaluation indicators were calculated using the entropy method. The entropy method is used to aggregate the weights of all tertiary evaluation indicators into the weights of secondary evaluation indicators; The entropy method is used to aggregate the weights of all secondary evaluation indicators into the weights of the primary evaluation indicators; The effectiveness of the formation in performing the current task is calculated based on the weights of the primary evaluation indicators.

6. The method for dynamic evaluation of game simulation effectiveness based on entropy value method according to claim 5, characterized in that, The entropy method includes: Calculate the weight of each indicator in each sample; The information entropy of a single indicator is calculated based on the aforementioned proportion. The weight of each indicator is calculated based on information entropy; Specifically, a sliding window mechanism is used to update the sample set used for calculation, so as to achieve dynamic updating of weights.

7. A dynamic evaluation system for game simulation effectiveness based on entropy method, characterized in that, include: The typical task and action set construction module is used to build typical task and action sets in game scenarios based on historical simulation data and through cluster analysis. The performance function design module is used to establish a multi-level evaluation index system for each type of typical task and to set performance functions for the final level index. The node information acquisition and task type identification module is used to acquire the status information and dynamic instructions of simulation nodes in real time during the simulation operation; based on the dynamic instructions and / or node status information, it matches the typical tasks and action sets to identify the task type currently being executed by each formation. The data-oriented acquisition module is used to load a corresponding multi-level evaluation index system subset for each formation according to the identified task type, and to collect data from relevant nodes in a targeted manner based on the multi-level evaluation index system subset; The final-level indicator performance value calculation module is used to calculate the performance value of mapping the data collected in the data-oriented acquisition module to the final-level indicator using the performance function. The formation performance calculation module is used to dynamically calculate the performance of each formation in performing the current task using the performance value of the final-level index as a sample and the entropy method.

Images