Intelligent judgment decision-making system in red and blue confrontation training

Abstract

The invention discloses an intelligent judgment decision-making system in red-blue confrontation training, and the system comprises a multi-source battlefield data collection module which is used for obtaining the military strength states of red and blue parties and a battle event in real time, and generating real-time battle situation data; the intelligent behavior simulation and confrontation judgment module is used for driving a blue military target to act based on the real-time battle situation data and outputting a hit part and damage grade judgment result; the multi-dimensional battlefield evaluation and redisk analysis module is used for carrying out battlefield damage statistics and capability evaluation and supporting two / three-dimensional situation playback and key event reproduction; and the closed-loop feedback and guidance intervention module is used for dynamically adjusting the confrontation intensity according to the evaluation result and the training target, and realizing real-time intervention of the blue army action strategy and closed-loop regulation of the training process. According to the embodiment of the invention, objectization, intelligentization and process of red-blue confrontation training judgment can be realized, and the fidelity, the evaluation accuracy and the group training efficiency of military training are improved.

Description

Technical Field

[0001] This invention belongs to the field of simulation decision-making technology, and in particular to an intelligent referee decision-making system for red-blue team training. Background Technology

[0002] In modern military training systems, red-blue force-on-force exercises are a key means of enhancing troops' combat capabilities. Traditional adversarial training decisions rely heavily on manual judgment and fixed rules, resulting in high subjectivity and low efficiency. While computer systems have been used for some adjudication with the development of simulation technology, existing systems typically have the following limitations: First, battlefield data collection dimensions are limited, making it difficult to construct high-fidelity real-time combat scenarios; second, behavioral simulation and adjudication models are rigid, and the intelligence of the blue force targets is insufficient, failing to realistically simulate complex and ever-changing adversarial environments; third, training evaluation is mostly focused on post-event debriefing, lacking dynamic adjustment and closed-loop feedback capabilities based on real-time data during training. This limits training effectiveness and makes it difficult to accurately assess and improve troops' combat effectiveness and responsiveness under near-real combat conditions. Summary of the Invention

[0003] The purpose of this invention is to provide an intelligent referee decision-making system for red-blue confrontation training to overcome the shortcomings of existing technologies. This system can achieve objectivity, intelligence, and process-orientation in red-blue confrontation training decisions, thereby improving the realism of military training, the accuracy of assessment, and the efficiency of training organization.

[0004] One embodiment of this application provides an intelligent referee decision-making system for red-blue team training, the system comprising: The multi-source battlefield data acquisition module is used to acquire the real-time troop status and combat events of both the red and blue sides. It collects location information, equipment status and ammunition consumption data through sensor arrays deployed on live-fire equipment and target terminals, and generates real-time combat situation data by combining digital twin environment mapping. The intelligent behavior simulation and adversarial adjudication module is used to drive the Blue Force target actions and calculate the strike effect based on real-time combat situation data. By integrating ballistic models, damage models and autonomous combat rules, it simulates the Blue Force fire strike process and outputs the hit location and damage level judgment results. The multi-dimensional battlefield assessment and debriefing analysis module is used to conduct battle damage statistics and capability assessment based on damage level and combat events. It generates combat effectiveness analysis reports by combining preset assessment indicator systems and supports time-series-based two / three-dimensional situation replay and key event reproduction. The closed-loop feedback and guidance intervention module is used to dynamically adjust the intensity of the confrontation based on the evaluation results and training objectives. It realizes real-time intervention in the blue team's action strategy and closed-loop control of the training process through the human-in-the-loop control interface and intelligent algorithm-driven mode.

[0005] Optionally, the multi-source battlefield data acquisition module is specifically used for: By deploying multi-source sensor arrays on Red Army individual soldier equipment and Blue Army target terminals, raw data including position coordinates, attitude orientation, equipment operating status and ammunition consumption are collected in real time to generate raw sensor data streams. The raw sensor data stream is preprocessed and fused. The Kalman filter algorithm is used to eliminate noise and compensate for data delay, generating calibrated multi-source state data. The calibrated multi-source state data is input into the digital twin environment, and the red and blue forces in the physical world are mapped one-to-one with the virtual model through the entity mapping algorithm to generate a digital twin entity state mapping table. Based on a digital twin entity state mapping table, and combined with real-time combat event triggers to detect fire strikes and movement behavior, the virtual battlefield situation is dynamically updated, ultimately generating real-time combat situation data.

[0006] Optionally, the intelligent behavior simulation and adversarial adjudication module is specifically used for: Analyze the blue force's troop positions, red force's target distribution, and environmental information from real-time combat situation data, input them into the autonomous combat rule model to make behavioral decisions, and generate blue force target action control instructions; According to the Blue Army's target action control instructions, the target is driven to perform concealment, maneuvering or attack actions, while the ballistic model is integrated to calculate the flight trajectory of digital munitions and generate simulated ballistic data. The collision rendezvous is calculated based on simulated ballistic data, and the collision point is detected by combining terrain occlusion and target attitude. The hit location and hit timestamp are output. The damage model is invoked based on the hit location, and the damage effect is calculated by combining the ammunition type and target protection data. Finally, the damage level determination result is output.

[0007] Optionally, the multi-dimensional battlefield assessment and debriefing analysis module is specifically used for: Collect damage level assessment results and combat event records, classify and aggregate data according to the red and blue force organizational structure, and generate a battle damage statistics set and an ammunition consumption summary table. Input the battle damage statistics into the preset evaluation index system, calculate the combat effectiveness index score, and the combat effectiveness index includes at least the survival rate, hit rate and mission completion rate to generate preliminary evaluation results; Based on the preliminary assessment results and time-series combat data, an operational effectiveness analysis report is automatically generated, which includes chart-based comparisons of combat damage and analysis of key events. By using a time-series database to replay two- or three-dimensional situations, and by using event triggers to reproduce key combat scenarios, an interactive debriefing and analysis interface is finally output.

[0008] Optionally, the closed-loop feedback and guidance intervention module is specifically used for: Analyze the evaluation results in the combat effectiveness analysis report, compare them with the pre-set training target identification deviation, and generate the need to adjust the intensity of the confrontation. Based on the need to adjust the intensity of the confrontation, a dynamic difficulty control strategy is designed. The strategy includes adjusting the hit rate, reconnaissance range or troop deployment of the blue force, and generating a set of control parameters. The human-in-the-loop control interface allows the director to manually intervene in the Blue Team's actions, or automatically send control parameter sets to the Blue Team's control terminal through the intelligent algorithm-driven mode to generate real-time control commands; It executes real-time control commands and monitors the training process, iteratively optimizes control parameters based on real-time feedback data, and achieves closed-loop control of the training process.

[0009] Another embodiment of this application provides an intelligent referee decision-making method in red-blue team training, the method comprising: Real-time acquisition of the troop status and combat events of both the red and blue forces; collection of location information, equipment status and ammunition consumption data by sensor arrays deployed on live-fire equipment and target terminals; and generation of real-time combat situation data by combining digital twin environment mapping. Driven by real-time combat situation data, the system simulates the Blue Force's firepower strike process and calculates the strike effect by integrating ballistic models, damage models and autonomous combat rules, and outputs the hit location and damage level determination results. Based on damage levels and combat events, the system performs combat damage statistics and capability assessments, generates combat effectiveness analysis reports by combining preset assessment indicator systems, and supports time-series-based two / three-dimensional situational awareness replay and key event reproduction. The intensity of the confrontation is dynamically adjusted based on the evaluation results and training objectives. Real-time intervention in the Blue Army's action strategy and closed-loop control of the training process are achieved through the human-in-the-loop control interface and intelligent algorithm-driven mode.

[0010] Another embodiment of this application provides a storage medium storing a computer program, wherein the computer program is configured to execute the method described in any of the preceding claims when running.

[0011] Another embodiment of this application provides an electronic device including a memory and a processor, wherein the memory stores a computer program and the processor is configured to run the computer program to perform the method described in any of the preceding claims.

[0012] Compared with existing technologies, this invention provides an intelligent referee decision-making system for red-blue confrontation training. The system includes: a multi-source battlefield data acquisition module for real-time acquisition of the troop status and combat events of both red and blue sides, generating real-time combat situation data; an intelligent behavior simulation and confrontation adjudication module for driving the blue force's target actions based on real-time combat situation data, outputting the hit location and damage level judgment results; a multi-dimensional battlefield assessment and debriefing analysis module for conducting battle damage statistics and capability assessment, and supporting two / three-dimensional situation playback and key event reproduction; and a closed-loop feedback and guidance intervention module for dynamically adjusting the confrontation intensity according to the assessment results and training objectives, realizing real-time intervention in the blue force's action strategy and closed-loop control of the training process, thereby achieving objectivity, intelligence, and process-orientation of red-blue confrontation training adjudication, improving the realism of military training, the accuracy of assessment, and the efficiency of training organization. Attached Figure Description

[0013] Figure 1 A structural block diagram of an intelligent referee decision-making system in red-blue team training provided by an embodiment of the present invention; Figure 2 This is a flowchart illustrating an intelligent referee decision-making method in red-blue team training, as provided in an embodiment of the present invention. Detailed Implementation

[0014] The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present invention, and should not be construed as limiting the present invention.

[0015] See Figure 1 The present invention provides an intelligent referee decision-making system for red-blue team training, the system comprising: The multi-source battlefield data acquisition module 101 is used to acquire the troop status and combat events of both the red and blue sides in real time. It collects location information, equipment status and ammunition consumption data through sensor arrays deployed on live-fire equipment and target terminals, and generates real-time combat situation data by combining digital twin environment mapping. Specifically, the multi-source battlefield data acquisition module is used for: By deploying multi-source sensor arrays on Red Army individual soldier equipment and Blue Army target terminals, raw data including position coordinates, attitude orientation, equipment operating status and ammunition consumption are collected in real time to generate raw sensor data streams. This step is fundamental to multi-source battlefield data acquisition. By deploying diverse sensors on physical equipment and targets, it comprehensively captures the real-time operational status of both the red and blue teams, providing raw material for subsequent data processing. The specific implementation is as follows: The deployment of the sensor array needs to cover two types of objects: Red Army soldiers and Blue Army targets, and appropriate sensors should be configured according to the characteristics of different objects. The sensor configuration of Red Army individual soldier equipment: Each soldier is equipped with a multi-mode positioning module (integrating GPS and Beidou, positioning accuracy ±1m, update frequency 1Hz), a six-axis attitude sensor (measurement range 0-360°, attitude accuracy ±0.5°, sampling frequency 10Hz), an ammunition counter (integrated into the gun magazine, automatically decrementing the count after each round fired, counting accuracy ±0), and equipment status monitoring sensors (detecting whether the gun is in a ready-to-fire state, whether the individual soldier's protective equipment is intact, status resolution "normal / abnormal"). For example, the real-time sensor data collected by Red Army soldier A (equipment ID: R-Soldier-001) is as follows: 10:00:00, position coordinates (112.3378°E, 37.3612°N, 50.2m elevation), attitude orientation (pitch angle 0°, roll angle 0°, yaw angle 90°, i.e. facing due east), ammunition consumption (28 rounds of 5.8mm rifle ammunition remaining, initial 30 rounds, 2 rounds fired), equipment status (weapon ready to fire, body armor intact).

[0016] The sensor configuration of the Blue Army target terminal is as follows: Each target (such as a fixed rifle hand target or a mobile armored target) is equipped with a GPS positioning module (accuracy ±0.5m, update frequency 2Hz), an attitude sensor (detects the target's visibility status, resolution "visible target / concealed target", response time <100ms), a hit sensor (distributed on the target's head, torso, limbs, etc., to detect live bullet hit signals, hit determination accuracy ±1cm), and an equipment operation status sensor (detects whether the target's motor and communication module are normal, status resolution "normal / fault"). For example, the real-time sensor data collected by Blue Army target B (equipment ID: B-Target-005) is as follows: 10:00:00, position coordinates (112.3385°E, 37.3608°N, 50.0m elevation), attitude status (visible target status, elevation angle 0°), hit status (no hit signal), and equipment status (motor normal, communication normal).

[0017] The raw sensor data stream is a time-series collection of data acquired by all sensors, organized in the format of "timestamp-device type-device ID-parameter name-parameter value-data quality identifier". Each data point is accompanied by a timestamp (accurate to milliseconds, such as 10:00:00.123) and a data quality identifier ("valid / invalid", such as invalid when the GPS signal is weak). For example, a raw data stream record is as follows: "10:00:00.123, Red Army soldier, R-Soldier-001, position coordinates, (112.3378,37.3612,50.2), valid; 10:00:00.123, Blue Army target, B-Target-005, attitude status, target visible, valid; 10:00:00.234, Red Army soldier, R-Soldier-001, ammunition remaining, 28 rounds, valid." The data stream needs to be transmitted to the data processing server in real time (transmission latency < 200ms) to ensure that no original information is lost.

[0018] The raw sensor data stream is preprocessed and fused. The Kalman filter algorithm is used to eliminate noise and compensate for data delay, generating calibrated multi-source state data. This step improves the reliability and timeliness of the original data through data optimization, providing high-quality input for subsequent digital twin mapping. The specific implementation is as follows: Preprocessing targets outliers and invalid data in the original data stream using a "threshold removal + interpolation completion" strategy. Outlier removal: Set reasonable ranges for each parameter (e.g., location coordinates must be within the training ground boundary; Red Army individual soldier positions must be within the range of 112.335°E-112.340°E and 37.360°N-37.365°N at the Guanting training ground; remaining ammunition ≥ 0). Data exceeding these ranges are marked as outliers and removed. For example, if a soldier's GPS data jumps to 112.500°E (outside the training ground), it is judged as outlier and removed. Interpolation completion: For the removed abnormal data or temporarily missing data (such as 50ms of sensor offline time), linear interpolation is used for completion. For example, if the position of Red Army soldier A at 10:00:00.123 is (112.3378, 37.3612) and the position at 10:00:00.234 is (112.3379, 37.3613), then the missing position at 10:00:00.178 is completed as (112.33785, 37.36125).

[0019] The fusion processing employs a Kalman filter algorithm, which is primarily used to eliminate sensor noise (such as random jumps in GPS positioning and minute fluctuations in attitude sensors) and compensate for data latency (such as a 100ms sensor data transmission delay). Noise cancellation: Taking GPS location data as an example, the raw data may exhibit jumps such as "112.3378→112.3385→112.3379" due to signal interference. Through the prediction-update loop of Kalman filtering (the prediction step estimates the next position based on the motion model, and the update step combines the measurement value for correction), a smoothed position sequence "112.3378→112.3379→112.3379" is output. The filtering parameters are set to process noise covariance Q=0.01 (reflecting the uncertainty of the motion model) and measurement noise covariance R=0.1 (reflecting the sensor measurement error), ensuring that the filtering effect balances smoothness and response speed. Delay compensation: To address data transmission delays, the actual state at the current moment is predicted by using the movement trend of historical data. For example, if the data collected by the sensor at 10:00:00.000 is delayed until 10:00:00.100, based on the movement speed of this data (e.g., 0.1m / s eastward), the actual position at 10:00:00.100 is predicted to be the original position + 0.01m (0.1m / s × 0.1s), thus achieving delay compensation.

[0020] The calibrated multi-source status data needs to integrate all optimized parameters and be output in the format of "Device ID-Timestamp-Location Coordinates (after calibration)-Attitude Orientation (after calibration)-Equipment Status (after calibration)-Ammunition Remaining (after calibration)". For example, the calibration data for Red Army soldier A is: "R-Soldier-001, 10:00:00.123, (112.3378,37.3612,50.2), (0°,0°,90°), Weapon Ready to Fire / Bodyguards Intact, 28 Rounds"; the calibration data for Blue Army target B is: "B-Target-005, 10:00:00.123, (112.3385,37.3608,50.0), Target Displayed, Motor Normal / Communication Normal, No Ammunition Attributes (Target Has No Ammunition)". The data must ensure the consistency of each parameter (such as position and attitude matching motion logic) to lay the foundation for digital twin mapping.

[0021] The calibrated multi-source state data is input into the digital twin environment, and the red and blue forces in the physical world are mapped one-to-one with the virtual model through the entity mapping algorithm to generate a digital twin entity state mapping table. This step uses digital twin technology to establish a connection between the physical and virtual worlds, achieving a digital mapping of battlefield conditions. The specific implementation is as follows: The digital twin environment needs to preload the geographic information model of the training ground (such as the terrain and feature model of the Guanting training ground, with an accuracy of 1m) and the three-dimensional physical models of the red and blue forces (such as the digital human body model of the red army soldier and the armored / infantry target model of the blue army). The model needs to include parameter interfaces (position, attitude, status, ammunition) corresponding to the physical equipment and support real-time parameter updates.

[0022] The entity mapping algorithm employs a "unique identifier + parameter association" strategy to achieve a precise correspondence between physical devices and virtual models. Unique Identifier Matching: A unique hardware ID (e.g., R-Soldier-001, B-Target-005) is assigned to each physical device (Red Army individual equipment, Blue Army target terminal), and a unique virtual ID (e.g., VR-Soldier-001, VB-Target-005) is assigned to the virtual model in the digital twin environment. The association is established through an ID mapping table. For example, the hardware ID “R-Soldier-001” corresponds to the virtual ID “VR-Soldier-001”, and the hardware ID “B-Target-005” corresponds to the virtual ID “VB-Target-005”. Parameter association mapping: Each parameter in the calibrated multi-source state data is bound to the corresponding attribute of the virtual model according to the preset mapping rules. For example, the physical device's "position coordinates (x,y,z)" is mapped to the virtual model's "world coordinates (x,y,z)", "attitude orientation (pitch angle, roll angle, yaw angle)" is mapped to the virtual model's "attitude angles (Pitch, Roll, Yaw)", "equipment status - weapon ready to fire" is mapped to the virtual model's "weapon status - Ready", and "ammunition remaining 28 rounds" is mapped to the virtual model's "ammunition quantity - 28".

[0023] The digital twin entity status mapping table needs to integrate the above-mentioned relationships and include fields such as "physical device type", "physical hardware ID", "virtual model type", "virtual model ID", "parameter mapping list", and "mapping status". For example, a mapping record is as follows: "Red Army individual soldier equipment, R-Soldier-001, Red Army infantry digital model, VR-Soldier-001, position (x,y,z) → world coordinates (x,y,z), attitude (Pitch,Roll,Yaw) → attitude angle (Pitch,Roll,Yaw), weapon status → weapon status, ammunition remaining → ammunition quantity, mapping is normal; Blue Army target terminal, B-Target-005, Blue Army fixed infantry target model, VB-Target-005, position (x,y,z) → world coordinates (x,y,z), attitude status → visible / hidden status, device status → running status, mapping is normal." The mapping table needs to be updated in real time (the update frequency is consistent with the sensor acquisition frequency, 1Hz). If a physical device is offline (e.g., the sensor has no data), the mapping status is marked as "abnormal" and the parameter update of the virtual model is paused.

[0024] Based on a digital twin entity state mapping table, and combined with real-time combat event triggers to detect fire strikes and movement behavior, the virtual battlefield situation is dynamically updated, ultimately generating real-time combat situation data.

[0025] This step transforms static mapping into a dynamic battlefield scenario through event detection and situational updates, providing real-time situational support for subsequent adversarial adjudication. The specific implementation is as follows: Real-time combat event triggers are the core module for detecting key battlefield behaviors, and include two types of triggering logic: fire strike detection and movement behavior detection. Firepower Detection: Through multi-source data collaborative judgment, when the following conditions are met simultaneously: "the ammunition counter of the Red Army's individual equipment decreases (e.g., from 28 rounds to 27 rounds, judged as ammunition firing)", "the hit sensor of the Blue Army's target terminal is triggered (e.g., a live impact signal is detected, judged as a hit)", and "the positions and attitudes of both satisfy ballistic logic (e.g., the angle between the Red Army's orientation and the Blue Army's position is <5°, and the distance is within the effective range)", a "firepower strike event" is triggered, and the event timestamp (e.g., 10:00:02.345), the striker ID (R-Soldier-001), the attacked party ID (B-Target-005), and the strike type (live hit) are recorded. Motion behavior detection: Based on the continuous changes in the position of physical devices, if the position distance of a physical device (such as Red Army soldier A) is greater than 0.1m within two consecutive acquisition cycles (excluding measurement errors), a "motion behavior event" is triggered, recording the direction of movement (such as due east), the speed of movement (such as 0.1m / s), the starting position, and the ending position; if the Blue Army target terminal switches from the "hidden target" state to the "revealed target" state, a "target reveal / hide event" is triggered, recording the state switching time and the current position.

[0026] Dynamically updating the virtual battlefield situation requires adjusting the virtual model state synchronously based on the aforementioned events and mapping table. After a fire strike event is triggered, the status of the virtual model of the attacked party is updated (e.g., the Blue Army target VB-Target-005 changes from "intact" to "hit", and displays the hit area effect). After a motion behavior event is triggered, the position of the virtual model of the moving party is updated (e.g., Red Army VR-Soldier-001 moves from (112.3378, 37.3612) to (112.3379, 37.3613)). After the target visibility event is triggered, update the visibility state of the target virtual model (e.g., VB-Target-005 changes from transparent to visible).

[0027] Real-time combat situation data is a dynamic snapshot of the virtual battlefield, including "situation timestamps," "summary of Red and Blue force troop status," "list of real-time combat events," and "battlefield environment status." The situation data is divided into four parts. For example, the situation data at a certain moment is as follows: "Situation timestamp: 10:00:02.345; Red Army force status: R-Soldier-001 (location 112.3379, 37.3613, ammunition 27 rounds, weapon ready to fire), R-Vehicle-001 (armored vehicle, location 112.3380, 37.3615, status normal); Blue Army force status: B-Target-005 (location 112.3385, 37.3608, status hit), B-Target-006 (location 112.3386, 37.3609, status target displayed); Real-time combat event: 10:00:02.345, fire strike, striker R-Soldier-001, attacked B-Target-005; Battlefield environment status: clear weather, wind force 0, visibility 10km." Situational data must be generated and stored at a frequency of 1 Hz to ensure that it can be used for retrospective analysis and adjudication in the future.

[0028] The intelligent behavior simulation and adversarial adjudication module 102 is used to drive the Blue Army target action based on real-time combat situation data and calculate the strike effect. By integrating ballistic model, damage model and autonomous combat rules, it simulates the Blue Army fire strike process and outputs the hit location and damage level judgment results. Specifically, the intelligent behavior simulation and adversarial adjudication module is used for: Analyze the blue force's troop positions, red force's target distribution, and environmental information from real-time combat situation data, input them into the autonomous combat rule model to make behavioral decisions, and generate blue force target action control instructions; This step is the core decision-making process in intelligent behavioral simulation. By extracting key real-time battlefield information and combining it with tactical rules, it determines the specific actions of the opposing force target. The specific implementation is as follows: Real-time combat situation data contains three core types of information, which need to be broken down and structured according to the dimensions of "troop strength-target-environment" during analysis: Blue Force Positions: Extract the real-time coordinates (accurate to the meter level), current attitude (visible / hidden target, elevation / yaw angle), and combat status (whether it can be attacked, remaining simulated ammunition) of all Blue Force targets from the situational data. For example, the analysis result of Blue Force Target No. 1 (type: fixed rifle hand target, ID: B-T001) is "position (112.3385°E, 37.3608°N, 50.0m elevation), attitude (visible target status, elevation angle 0°, yaw angle 180°, i.e. facing due west), combat status (can be attacked, simulated 30 rounds of 5.8mm rifle ammunition remaining)"; Red Team Target Distribution: Extract the location, quantity, and threat level of individual Red Army soldiers / equipment. Threat level is divided into levels 1-5 (1-5, with level 5 being the highest) based on the proximity to the Blue Team and the strength of the equipment's firepower. For example, the analysis results for two Red Team soldiers are "Red 1 (ID: R-S001, location 112.3378°E, 37.3612°N, equipped with a 5.8mm rifle, threat level 4) and Red 2 (ID: R-S002, location 112.3390°E, 37.3605°N, equipped with a pistol, threat level 2)". Environmental information: Extract the terrain type (flat / hilly / built area), meteorological data (wind speed 0.5m / s, wind direction southeast, visibility 10km) and obstacle distribution of the training ground (such as a 1.5m high wall at 112.3382°E, 37.3610°N). This information directly affects the Blue Army's action decisions (such as bypassing obstacles and adjusting the shooting angle according to the wind direction).

[0029] The autonomous engagement rules model needs to integrate the Blue Force's tactical principles (referencing the system's preset "strong enemy" combat rules) with the real-time situation. The decision-making logic consists of three steps: Target selection: Prioritize targets with high threat levels and no terrain obstruction. For example, Red 1 has a threat level of 4 and there are no obstacles between it and Blue 1 target. Red 2 has a threat level of 2 and is obstructed by a low wall. Therefore, Red 1 is the priority target. Action type determination: If the target is within the effective range (blue target 1 simulates a rifle with an effective range of 400m, red target 1 is 45m away, within the range), it is determined as "attack action"; if the target is beyond the range but can be reached by maneuver (e.g., red target 2 is 60m away, and enters the range after maneuvering 15m), it is determined as "maneuver + attack action"; if the target has strong cover and low threat, it is determined as "target evasion". Generate control commands: For "strike operations", the command must include the type of ammunition (simulating 5.8mm rifle rounds), the number of rounds fired (2 rounds to ensure a hit probability), and the firing azimuth angle (calculate the azimuth angle of Red 1 relative to Blue 1 as 275°, matching the target yaw angle); For "maneuver operations", the command must include the maneuver target point (112.3382°E, 37.3610°N, avoiding low walls), the maneuver speed (2m / s, meeting the target's upper limit of maneuverability of 3m / s), and the maneuver duration (7.5 seconds, corresponding to a maneuver distance of 15m).

[0030] The final generated blue team target action control instructions must clearly specify "target ID - action type - specific parameters - execution sequence", for example, "B-T001, strike action, ammunition type: 5.8mm rifle rounds, quantity: 2 rounds, azimuth angle: 275°, execution sequence: 10:00:05.000-10:00:05.100 (first round 10:00:05.000, second round 10:00:05.100)", to ensure that the target can execute the instructions accurately.

[0031] According to the Blue Army's target action control instructions, the target is driven to perform concealment, maneuvering or attack actions, while the ballistic model is integrated to calculate the flight trajectory of digital munitions and generate simulated ballistic data. This step, through hardware drivers and ballistic simulation, translates decision commands into target physical actions and virtual ammunition trajectories, as specifically implemented below: The blue team target must invoke the corresponding hardware control module based on the command type to execute actions, ensuring that the action response latency is less than 100ms. Concealment and Revealment Actions: In response to the "Concealment Target Avoidance" command, the target is driven by a built-in motor to perform a cyclical action according to the "Concealment Time (e.g., 5 seconds)" and "Revealment Time (e.g., 10 seconds)" set in the command. For example, if the command requires "10:00:06.000-10:00:11.000 Concealment Target", the target will trigger the tilting action (from upright 30° to horizontal 0°) at 10:00:06.000 and the repositioning action (returning to upright) at 10:00:11.000. The action accuracy is ±1°, ensuring that the concealment and revealing state is synchronized with the virtual situation. Maneuvering Actions: In response to the "maneuvering + strike" command, the maneuvering target (such as a wheeled armored target) moves along a preset route (from the initial position to the target point) at the commanded speed (2m / s) using its drive wheels. The real-time position is updated every 500ms and fed back to the digital twin environment. For example, target Blue 2 (ID: B-T002) moves from (112.3395°E, 37.3600°N) to (112.3385°E, 37.3605°N). During the process, the position is (112.3392°E, 37.3602°N) at 10:00:05.500 and (112.3389°E, 37.3603°N) at 10:00:06.000, ensuring that the maneuvering trajectory is consistent with the command. Strike Action: In response to the "Strike Action" command, the target outputs a strike signal through a simulated firing module (such as a laser emitter, used to trigger hit detection for individual Red Army soldiers) and records the firing timestamp. For example, target Blue 1 triggers laser firing twice at 10:00:05.000 and 10:00:05.100 to simulate ammunition firing action, with the firing frequency consistent with the command setting.

[0032] The integrated ballistic model needs to consider four types of physical factors to ensure the accuracy of trajectory calculation: muzzle velocity: set according to the simulated ammunition type, such as 920m / s for 5.8mm rifle bullets and 850m / s for 12.7mm machine gun bullets, with a muzzle velocity error of ±10m / s; Terrain data: Import the elevation model of the training ground (e.g., the elevation of target 1 is 50.0m, the elevation of target 1 is 50.2m, and the relative elevation difference is 0.2m), and calculate the elevation correction of the trajectory (original azimuth 275°, corrected to 275°, elevation +0.2° to offset the elevation difference). Meteorological data: A wind speed of 0.5 m / s (southeast direction) will cause the projectile to deviate laterally. Calculate the deviation (flight time 0.05 seconds, deviation = wind speed × flight time = 0.025 m, i.e. 2.5 cm), and correct the trajectory endpoint coordinates. Effective range: If the projectile's flight distance exceeds the effective range (e.g., the effective range of a simulated 12.7mm machine gun bullet is 1000m, and its kinetic energy is insufficient after flying 1200m), it is judged as "ballistic attenuation", and the trajectory endpoint is set at the effective range.

[0033] The simulated ballistic data is the time-series coordinates of the projectile's entire trajectory from launch to impact (or landing), generated in the format of "timestamp (0.01 second interval) - x coordinate (E) - y coordinate (N) - z coordinate (elevation)". For example, the trajectory data of a 5.8mm rifle bullet fired from target blue 1 is: "10:00:05.000, 112.3385°E, 37.3608°N, 50.0m; 10:00:05.010, 112.3384°E, 37.3608°N, 50.05m; ... 10:00:05.050, 112.3379°E, 37.3612°N, 50.2m (hitting red 1 position)". The coordinates of each timestamp are adjusted for terrain and weather to ensure that the trajectory conforms to physical laws.

[0034] The collision rendezvous is calculated based on simulated ballistic data, and the collision point is detected by combining terrain occlusion and target attitude. The hit location and hit timestamp are output. This step uses spatial collision detection to accurately determine whether the projectile hit the target and the point of impact. The specific implementation is as follows: The calculation of intersection hits requires verification in three steps to ensure no missed or incorrect hits: Trajectory and target spatial overlap detection: Extract the endpoint coordinates of the simulated ballistic data (e.g., 112.3379°E, 37.3612°N, 50.2m) and the spatial range of the Red Team target (Red 1 soldier's human model dimensions: height 1.75m, shoulder width 0.5m, front-to-back diameter 0.3m, center point coordinates 112.3378°E, 37.3612°N, 50.875m), and calculate the spatial distance between the two (horizontal distance 0.011km ≈ 11m? No, the actual horizontal distance is √[(112.3379-112.3378)²). +(37.3612-37.3612)²]×111319.9m / °≈11.13m? No, it should be a small range of coordinate differences. For example, the coordinates of red 1 are (112.337800°E, 37.361200°N), and the trajectory endpoint is (112.337805°E, 37.361203°N). The horizontal distance is ≈√[(0.000005×111319.9)²+(0.000003×111319.9)²]≈0.67m. Within a range of 0.5m shoulder width, the space is considered to overlap. Terrain occlusion verification: Call the terrain model of the digital twin environment to check whether the line connecting the starting point (blue 1) and the ending point (red 1) of the trajectory is occluded by obstacles. For example, if the height of the line connecting the two points is higher than the low wall (1.5m) and there are no other buildings, it is determined that there is no occlusion; if the line passes through the low wall (such as the line connecting the red 2 position and the blue 1 passing through the 1.5m low wall), it is determined that "occlusion hit is invalid". Target posture adaptation detection: The red team's target posture (standing / lying down / crawling) will change the range of human body space. For example, if red 1 is in a standing posture (height 1.75m, z coordinate range 50.0-51.75m), the ballistic endpoint z coordinate of 50.2m is within this range; if red 1 is in a lying posture (height 0.5m, z coordinate range 50.0-50.5m), it is necessary to verify whether the z coordinate is within this range to avoid misjudgment caused by posture.

[0035] Collision point detection needs to be combined with the division of human / equipment parts of the red team target: Personnel Target: The human body is divided into five parts: head (z coordinate 51.25-51.75m), left chest (x-0.2 to x-0.05m, y±0.1m, z50.8-51.25m), right chest (x+0.05 to x+0.2m, y±0.1m, z50.8-51.25m), left abdomen (x-0.2 to x-0.05m, y±0.1m, z50.5-50.8m), and right abdomen (x+0.05 to x+0.2m, y±0.1m, z50.5-50.8m). For example, if the ballistic endpoint z coordinate is 50.2m and falls within the left abdomen area, it is determined that the hit was to the left abdomen. Equipment target: Divide the armored equipment into four areas: front (x+0.3 to x+0.5m), rear (x-0.3 to x-0.5m), left (y+0.2 to y+0.3m), and right (y-0.2 to y-0.3m). For example, if the ballistics hit the armored equipment at x+0.4m, it is judged as a hit to the front.

[0036] The hit timestamp is the first timestamp in the ballistic data that overlaps with the target space. For example, if the ballistic trajectory first reaches the left abdomen of Red 1 at 10:00:05.050, the hit timestamp is "10:00:05.050". The final output must include "Ballistic ID - Target ID - Hit Location - Hit Timestamp - Occlusion Status", for example, "Traj-001, R-S001, Left Abdomen, 10:00:05.050, No Occlusion (Valid Hit)".

[0037] The damage model is invoked based on the hit location, and the damage effect is calculated by combining the ammunition type and target protection data. Finally, the damage level determination result is output.

[0038] This step determines the final damage level by quantifying the destructive capability of the ammunition and the protective capability of the target. The specific implementation is as follows: The damage model needs to first define two types of core input parameters, whose values ​​are derived from the system's preset equipment database and real-time situation: Ammunition type and damage parameters: Different ammunition types have different destructive force fields. For example, the kinetic energy density of a 5.8mm rifle bullet (standard bullet) is 150J / cm² (kinetic energy upon impact / impact area), and its kill radius is 0.5m; the kinetic energy density of a 12.7mm machine gun bullet is 800J / cm², and its kill radius is 1.2m; the armor penetration thickness of an anti-tank missile is 800mm, and its destruction radius is 3m. These parameters need to be determined based on the actual characteristics of the ammunition. Target protection data: The protection capability of personnel targets is determined by the equipment worn. For example, ordinary bulletproof vests (protection level III) can withstand the kinetic energy density of 5.8mm ordinary bullets ≤120J / cm², and the protection will fail if it exceeds this value; heavy bulletproof vests (level IV) can withstand ≤200J / cm². The protection capability of equipment targets is determined by the armor thickness. For example, the front armor thickness of a light armored vehicle is 15mm, which can withstand 12.7mm machine gun bullets (armor penetration thickness 10mm), but cannot withstand anti-tank missiles (800mm).

[0039] The damage effect calculation involves two steps: first, determining whether the protection is effective, and then calculating the specific degree of damage. Determining the effectiveness of protection: Compare the kinetic energy density of the ammunition with the target's protection limit. For example, if Red 1 is wearing ordinary bulletproof vest (protection limit 120J / cm²) and is hit by a 5.8mm rifle bullet (kinetic energy density 150J / cm²), 150 > 120, so the protection is deemed ineffective; if Red 1 is wearing heavy bulletproof vest (200J / cm²), 150 < 200, so the protection is deemed effective, causing only surface contusions. Damage assessment: Personnel Target (Protection Failure): The severity of the impact is categorized based on the location of the hit: a hit to the head / heart (left chest) is "Death" (Level 5); a hit to the abdomen / lungs (right chest) is "Severe Wound" (Level 4); a hit to the limbs is "Moderate Wound" (Level 3); a superficial contusion is "Minor Wound" (Level 2); and no hit is "Unharmed" (Level 1). For example, a hit to the left abdomen of Red 1 (Protection Failure) is classified as "Severe Wound." Equipment Target (Protection Failure): The level is determined based on the distribution of critical components in the hit area. A hit to the front / engine compartment is "Destroyed" (Level 3), a hit to the side / ammunition compartment is "Damaged" (Level 2), a hit to the rear / non-critical parts is "Slightly Damaged" (Level 1), and no hit is "Intact" (Level 0). For example, a hit to the front of a light armored vehicle by a 12.7mm machine gun round (protection effective) is classified as "Intact"; a hit by an anti-tank missile (protection failure) is classified as "Destroyed".

[0040] The damage assessment result must include "Target ID - Target Type - Hit Location - Ammunition Type - Protection Status - Damage Level - Level Description", for example, "R-S001, Personnel, Left Abdomen, 5.8mm Rifle Round, Protection Failed, Damage Level 4 (Severe Wound), Description: Soft tissue injury to the left abdomen, incapacitated, requires evacuation for treatment"; "B-V001, Light Armored Vehicle, Front, 12.7mm Machine Gun Round, Protection Effective, Damage Level 0 (Intact), Description: Armor not penetrated, power system normal". The results must be synchronized in real time to the digital twin environment and command terminal to provide a basis for subsequent battlefield assessment.

[0041] The multi-dimensional battlefield assessment and debriefing analysis module 103 is used to conduct battle damage statistics and capability assessment based on damage level and combat events, generate combat effectiveness analysis reports in combination with the preset assessment indicator system, and support time-series-based two / three-dimensional situation replay and key event reproduction. Specifically, the multi-dimensional battlefield assessment and debriefing analysis module is used for: Collect damage level assessment results and combat event records, classify and aggregate data according to the red and blue force organizational structure, and generate a battle damage statistics set and an ammunition consumption summary table. This step is the foundational data processing stage for multi-dimensional evaluation. It involves systematically collecting preliminary adversarial decisions and battlefield events, aggregating them according to operational group logic, and providing structured input for subsequent effectiveness calculations. The specific implementation is as follows: Data collection needs to cover two core types of information, sourced from the output of the intelligent behavior simulation and adversarial adjudication module and the records of real-time combat event triggers: Damage level assessment results: include the "target ID-group-target type (personnel / equipment / target)-hit location-damage level-assessment time" of all combat units on both the Red and Blue sides. For example, the assessment result for a soldier in the 1st Platoon, 1st Company, 1st Battalion of the Red Army (ID: R-S101) is "R-S101, 1st Platoon, 1st Company, 1st Battalion of the Red Army, personnel, left abdomen, severe injury (level 4), 10:05:30"; the assessment result for the Blue Army's No. 1 target group (ID: B-T001) is "B-T001, Blue Army Defense Group 1, fixed rifle hand target, torso, destroyed (level 3), 10:06:15". It is necessary to ensure that each assessment result is associated with a unique group and timestamp.

[0042] The combat event log contains all key battlefield actions in the format of "Event ID - Event Type (Fire Strike / Maneuver / Target Reveal / Hide) - Participant ID - Event Time - Event Location - Related Parameters". For example, the fire strike event log is "E001, Fire Strike, Striker R-S101, Target B-T001, 10:06:15, (112.3385°E, 37.3608°N), Ammunition Type 5.8mm Rifle Round"; the maneuver event log is "E002, Maneuver, Participant R-S102, 10:04:20, From (112.3378°E, 37.3612°N) to (112.3382°E, 37.3609°N), Speed ​​2m / s".

[0043] Categorization and aggregation must follow the pre-defined organizational structure of the Red and Blue Armies (Red Army at the "battalion-company-platoon-squad" level, Blue Army at the "defense group-target type" level) to ensure data alignment with operational organization: The logic for aggregating battle damage statistics is as follows: At the Red Army level, the battle damage data for the 1st Platoon of the 1st Company of the 1st Battalion needs to summarize the damage status of all individual soldiers / equipment in the platoon. For example, "1st Platoon of the 1st Company of the 1st Battalion of the Red Army: Total number of personnel: 12, 2 seriously wounded, 3 slightly wounded, 7 intact; Total number of equipment: 3 (2 machine guns, 1 armored vehicle), 1 armored vehicle damaged, 2 machine guns intact." At the Blue Army level, the battle damage data for Defense Group 1 needs to summarize the damage status of all targets in the group. For example, "Defense Group 1 of the Blue Army: Total number of targets: 8 (5 fixed rifle targets, 3 mobile armored targets), 5 destroyed, 2 damaged, 1 intact." The dataset needs to be organized in the format of "Group ID - Target Type - Total Quantity - Quantity of Each Damage Level" to ensure that the battle damage of each group is quantifiable and traceable.

[0044] The aggregation logic of the ammunition consumption summary table is as follows: Consumption and remaining quantities are categorized and statistically analyzed by "Squad - Group - Ammunition Type". Red Army ammunition consumption needs to be differentiated by live ammunition consumption (e.g., 5.8mm rifle rounds, 12.7mm machine gun rounds), while Blue Army ammunition consumption needs to be differentiated by simulated ammunition consumption (e.g., simulated 5.8mm rifle rounds, simulated anti-tank missiles). For example, "Red Army 1st Battalion 1st Company, 5.8mm rifle rounds: initial 300 rounds, consumed 85 rounds, remaining 215 rounds; 12.7mm machine gun rounds: initial 100 rounds, consumed 20 rounds, remaining 80 rounds; Blue Army Defense Group 1, simulated 5.8mm rifle rounds: initial 200 rounds, consumed 120 rounds, remaining 80 rounds; simulated anti-tank missiles: initial 10 rounds, consumed 3 rounds, remaining 7 rounds." The summary table needs to indicate the effective range and damage characteristics of each ammunition type (e.g., 5.8mm rifle round effective range 400m) to provide ammunition type background for subsequent effectiveness analysis.

[0045] Input the battle damage statistics into the preset evaluation index system, calculate the combat effectiveness index score, and the combat effectiveness index includes at least the survival rate, hit rate and mission completion rate to generate preliminary evaluation results; This step transforms battle damage data into comparable performance scores through quantitative evaluation indicators, laying the foundation for subsequent analysis reports. The specific implementation is as follows: The pre-set evaluation index system needs to be combined with the training objectives of the Red-Blue confrontation exercise (such as "improving the accuracy of firepower strikes and the survivability of personnel in urban offensive and defensive operations of the Red Army"), and the calculation logic, weight, and scoring criteria (out of 100 points) of each index should be clearly defined. The core indexes are designed as follows: Survival rate: Reflects the survivability of a combat unit in combat. The calculation logic is "Survival rate = (Total number of units - Number of casualties / damage) / Total number of units × 100", with a weight of 40% (personnel / equipment survival is the core of training safety and sustained combat). The scoring criteria are "≥80% = 100 points, 60%-80% = 80 points, 40%-60% = 60 points, <40% = 40 points". For example, the 1st platoon of the 1st company of the 1st battalion of the Red Army has a total of 12 personnel, with 5 casualties. The survival rate is (12-5) / 12×100≈58.3%, corresponding to a score of 60 points; the total number of equipment is 3, with 1 damaged. The equipment survival rate is (3-1) / 3×100≈66.7%, corresponding to a score of 80 points. The final survival rate score of the platoon is "personnel survival rate score × 0.6 + equipment survival rate score × 0.4" (personnel has a higher weight than equipment) = 60×0.6 + 80×0.4 = 36 + 32 = 68 points.

[0046] Hit rate: Reflects the accuracy of firepower strikes. The calculation logic is "Hit rate = (Number of effective hits / Total number of shots) × 100", with a weighting of 30% (accurate strikes are key to reducing casualties). An effective hit must meet the requirement of "hitting the target and causing damage". The total number of shots includes live ammunition and simulated shooting. The scoring standard is "≥50% gets 100 points, 30%-50% gets 80 points, 10%-30% gets 60 points, <10% gets 40 points". For example, the Red Army's 1st Battalion, 1st Company, fired a total of 150 shots and had 48 effective hits, with a hit rate of 48 / 150 × 100 = 32%, corresponding to a score of 80 points; the Blue Army's Defense Group 1 fired 200 simulated shots and had 55 effective hits, with a hit rate of 55 / 200 × 100 = 27.5%, corresponding to a score of 60 points. The hit rate scores for Red and Blue need to be calculated separately for subsequent comparison.

[0047] Task Completion Rate: Reflects the achievement of training tasks. The calculation logic is "Task Completion Rate = (Number of Completed Tasks / Total Number of Tasks) × 100", with a weighting of 30% (training must revolve around the task objectives). Task types include "capturing positions, destroying designated targets, and covering teammates' advance", etc. The scoring criteria are "≥80% = 100 points, 60%-80% = 80 points, 40%-60% = 60 points, <40% = 40 points". For example, the Red Army has a total of 5 tasks in this training (capturing 2 positions and destroying 3 targets), and has completed 3 (capturing 1 position and destroying 2 targets). The task completion rate is 3 / 5 × 100 = 60%, corresponding to a score of 80 points. The Blue Army has no active tasks, only "defending designated areas". The task completion rate is calculated as "number of areas not lost / total number of defended areas". If all 3 areas are not lost, 100 points are awarded.

[0048] The preliminary assessment results need to integrate the scores and weights of each indicator to calculate a weighted total score (total score = survival rate score × 0.4 + hit rate score × 0.3 + mission completion score × 0.3). For example, the preliminary assessment results of the 1st Company of the 1st Battalion of the Red Army are: "Survival rate 68 points × 0.4 = 27.2, hit rate 80 points × 0.3 = 24, mission completion 80 points × 0.3 = 24, weighted total score 75.2 points, level: good (80-100 points is excellent, 60-80 points is good, <60 points is needing improvement)". At the same time, the reasons for the deviation of each indicator need to be noted, such as "Hit rate 32% is lower than the expected 40%, reason: the wind direction effect (wind speed 0.5m / s southeast) was not considered when some soldiers fired", to provide a basis for subsequent report analysis.

[0049] Based on the preliminary assessment results and time-series combat data, an operational effectiveness analysis report is automatically generated, which includes chart-based comparisons of combat damage and analysis of key events. This step presents the evaluation results through a structured report, interprets battlefield patterns using time-series data, and provides an intuitive basis for training summaries. The specific implementation is as follows: The operational effectiveness analysis report should include four parts: "Assessment Overview - Battle Loss Comparison Analysis - Key Incident Analysis - Problem Summary". Each part should combine data with battlefield logic to avoid simply piling up numerical values. Assessment Overview: This concise summary presents the core assessment results for both the Red and Blue teams, including "Duration of the engagement (2 hours, 10:00-12:00), Participating forces (Red Army 1st Battalion 1st Company 50 men 10 pieces of equipment, Blue Army defense group 8 targets), Weighted total score (Red Army 75.2 points, Good; Blue Army 88 points, Excellent), and core conclusions (Red Army's mission completion rate met the standard, but the hit rate needs improvement; Blue Army's defense effect was significant, but the accuracy of simulated shooting needs optimization), allowing readers to quickly grasp the overall assessment situation.

[0050] Chart-based battle damage comparison: Although tables are not used, the core information of the charts must be described in text to clearly present the differences between red and blue. For example, “Personnel / Target Damage Comparison: Red Army personnel casualties were 12 (24%), mainly concentrated in the 10:30-11:00 phase of capturing position 2; Blue Army target damage was 6 (75%), mainly concentrated in the 11:10-11:30 phase of Red Army fire coverage. In terms of damage trend, Red Army casualties accounted for 60% in the first hour, and dropped to 40% in the second hour due to tactical adjustments. Blue Army target damage accounted for 40% in the first hour, and rose to 60% in the second hour due to the destruction of defensive fortifications.”; “Ammunition Consumption Comparison: Red Army consumed 210 rounds of 5.8mm rifle ammunition (4.2 rounds per person) and 45 rounds of 12.7mm machine gun ammunition (4.5 rounds per machine gun). Blue Army simulated ammunition consumption was 180 rounds (22.5 rounds per target). Red Army ammunition utilization rate (effective hits / consumption) was 22.9%, while Blue Army's was 30.6%. Blue Army had a higher utilization rate because simulated shooting was more focused on high-threat targets.”

[0051] Key event analysis: Select 3-5 core events that affect the assessment results and interpret the causal relationship by combining time-series combat data (timestamps, situation, casualties). For example, “Event 1: At 10:30, the 2nd Platoon of the 1st Company of the Red Army attacked Position 2 (defended by 3 targets of the Blue Army). The platoon did not clear the flank machine gun target (B-T003) first, and directly advanced from the front, resulting in 3 people being hit by simulated fire from B-T003 (2 seriously injured). Only 1 target was destroyed, and the mission was not completed. This event caused the Red Army’s survival rate to drop to 58% in the first hour, and the mission completion rate to only 40%; Event 2: At 11:20, the Red Army adjusted its tactics. The 1st Platoon provided fire support (suppressing the Blue Army’s B-T005 and B-T006), the 2nd Platoon flanked and destroyed the flank targets, and the 3rd Platoon advanced from the front. Within 15 minutes, 2 targets were destroyed without any casualties. This event increased the Red Army’s survival rate to 75% in the second hour and the mission completion rate to 80%.” Each event needs to be marked with a timestamp, participants, actions, results, and impact on evaluation indicators (e.g., Event 1 caused the hit rate to decrease by 5%).

[0052] Summary of problems: Based on the preliminary assessment results and event analysis, the strengths and weaknesses of the training were identified, such as "strengths: the Red Army's later tactical adjustments were effective, and the mission completion rate increased from 40% to 80%; weaknesses: the early fire strikes did not take wind direction into account (0.5 m / s in the southeast direction caused bullets to deviate by 2-3 cm), the hit rate was 10% lower than expected, some soldiers lacked coordination awareness, and flank protection was insufficient." Preliminary improvement suggestions were also given (such as "subsequent training should increase shooting practice in windy conditions and strengthen squad coordination tactics").

[0053] By using a time-series database to replay two- or three-dimensional situations, and by using event triggers to reproduce key combat scenarios, an interactive debriefing and analysis interface is finally output.

[0054] This step, through dynamic playback and scenario reproduction, allows training personnel to intuitively trace the confrontation process and pinpoint tactical problems. The specific implementation is as follows: The time-series database stores "timestamp-red / blue force status-environment data" for the entire combat exercise, with a time resolution of 1 second per record. Each record includes "the position of each unit of the red / blue forces (accurate to 0.1m), attitude (visible / hidden target, standing / prone), damage status, and remaining ammunition" as well as "real-time terrain and weather data". For example, the time-series data for 10:30:00 is "Red Army 1st Company 2nd Platoon position (112.3380°E, 37.3605°N), standing attitude, 3 severely wounded; Blue Army B-T003 position (112.3382°E, 37.3607°N), visible target attitude, 15 rounds of ammunition remaining; wind direction southeast 0.5m / s, 1.2m high / low wall on position 2". The database needs to support fast data retrieval by time range (e.g., 10:30-10:40) or event ID (e.g., E001).

[0055] Two- or three-dimensional situational awareness replays need to be based on time-series data and presented in a "timeline progression" mode. The two replay formats each have their own emphasis: Two-dimensional situational replay: On the electronic map of the training ground (1:1000 scale), red and blue positions are marked with military symbols (blue dots for individual Red Army soldiers, red triangles for Blue Army targets). Colors distinguish damage status (green for intact, yellow for minor damage, and red for severe damage / destruction). The timeline can be manually dragged (1-second accuracy) or played automatically (speed adjustable from 1x to 16x). For example, replaying the period from 10:30 to 10:35 shows the Red Army 2nd Platoon advancing from (112.3378°E, 37.3610°N) towards position 2. At 10:32, target B-T003 appears and simulates firing. At 10:33, the icons for two Red Army soldiers in the 2nd Platoon turn red (severe damage). At 10:35, only one target is destroyed (B-T004, icon turns red). The two-dimensional replay focuses on the troop movement trajectory and the sequence of casualties.

[0056] 3D Situation Replay: Based on a digital twin environment, 3D models (terrain, features, and individual soldier / target 3D models) recreate realistic battlefield scenes. Individual soldier models can display equipment (rifles, body armor), and target models can display hit effects (a red flash at the hit site). The viewpoint can be switched (third-person global view, individual soldier follow view, target shooting view). For example, switching to the follow view of the soldier (R-S105) of the 2nd Platoon, 1st Company of the Red Army, you can see that at 10:32, B-T003 appears as a target behind a low wall, and the simulated bullet trajectory (green line) points from B-T003 to the left of R-S105. At 10:33, the R-S105 model falls to the ground (severely wounded). The 3D replay more intuitively presents tactical details (such as whether the flank target was not detected due to obstruction by the low wall).

[0057] Event triggers need to be associated with key event IDs to enable the function of "clicking an event to replay the scene". For example, if you click "Event 1 (10:30 Attack on Position 2)" on the review interface, the system will automatically locate the time series data 10:30:00 and trigger the two- or three-dimensional replay from that time point. At the same time, the event details panel (participants, actions, casualties, ammunition consumption) will pop up. It supports pausing (to view the static situation), fast forward (skip the period without battle damage at 16x speed), and rewind (to rewind the pre-battle deployment).

[0058] An interactive debriefing interface should include a "timeline control area - situation display area - event list area - data details area": ​​The timeline control area has start / pause / fast forward / rewind buttons and a timeline; the situation display area allows switching between 2D and 3D views; the event list area lists all key events in chronological order (events highlighted in red that affect the assessment results); in the data details area, clicking on any unit (such as R-S105, B-T003) allows viewing its entire status (location, ammunition, and time of damage). For example, clicking on B-T003 displays "ID: B-T003, type: mobile machine gun target, deployment location: flank of position 2, simulated firing count: 30, effective hits: 8, destroyed by Red Army R-S108 at 11:25," helping training personnel trace the combat process of individual units and identify defensive vulnerabilities or offensive advantages.

[0059] The closed-loop feedback and guidance intervention module 104 is used to dynamically adjust the intensity of the confrontation based on the evaluation results and training objectives. It realizes real-time intervention and closed-loop control of the blue team's action strategy and training process through the human-in-the-loop control interface and intelligent algorithm-driven mode.

[0060] Specifically, the closed-loop feedback and guidance intervention module is used for: Analyze the evaluation results in the combat effectiveness analysis report, compare them with the pre-set training target identification deviation, and generate the need to adjust the intensity of the confrontation. This step is the starting point of the closed-loop feedback. By deeply interpreting the preliminary evaluation report, the gap between the training effect and the target is identified, and the direction for adjusting the intensity of subsequent adversarial training is clarified. The specific implementation is as follows: The combat effectiveness analysis report contains core assessment data for both the Red and Blue forces. Analysis must focus on key indicators strongly correlated with training objectives. These indicators must correspond one-to-one with pre-set training targets (set based on the needs of live-fire tactical training at battalion and company levels, such as "Red Army firepower hit rate ≥40%, personnel survival rate ≥70%, mission completion rate ≥80%; Blue Army defensive effectiveness ≥85%, simulated firing accuracy ≥60%)." For example, if the report shows the Red Army's assessment results as "hit rate 32%, survival rate 68%, mission completion rate 75%; Blue Army defensive effectiveness 90%, simulated firing accuracy 65%", analysis must compare the target values ​​with the actual values ​​one by one and calculate the deviation. Red Army hit rate deviation: Actual 32% - Target 40% = -8% (Negative deviation indicates failure to meet the target, and the Red Army hit rate needs to be improved, which can be achieved by reducing the intensity of the Blue Army's confrontation). Red Army survival rate deviation: Actual 68% - Target 70% = -2% (close to the target, minor adjustments are needed); Red Army mission completion deviation: Actual 75% - Target 80% = -5% (needs slight improvement, which can be achieved by optimizing the Blue Army's defensive deployment); Blue Army simulated shooting accuracy deviation: Actual 65% - Target 60% = +5% (If it exceeds the target, the Blue Army's shooting accuracy can be appropriately reduced to reduce the Red Army's survival pressure).

[0061] Deviation identification requires analysis of the causes of deviations in conjunction with battlefield events to avoid blind adjustments. For example, the core reason for the Red Army's low hit rate was that "the Blue Army's No. 1 mobile machine gun target (B-T003) had a simulated shooting hit rate of 65% and its reconnaissance range covered the Red Army's attack route (500m), causing the Red Army to frequently dodge shots and unable to aim steadily"; the Blue Army's high defensive effectiveness was due to "the No. 3 fixed target group (B-T005-007) being deployed in a narrow passage that the Red Army had to pass through, forming crossfire."

[0062] Based on the deviations and their causes, adjustments to the intensity of the confrontation are generated. These adjustments must clearly define the "object of adjustment (a specific type of target for the Blue Army / overall defense), the direction of adjustment (enhancement / weakening), and the purpose of adjustment (improving a specific indicator for the Red Army)." For example: "1. Adjust the Blue Army target B-T003: reduce the simulated shooting hit rate (from 65% → 55%) and narrow the reconnaissance range (from 500m → 400m) to reduce shooting interference during the Red Army's advance and improve the Red Army's hit rate; 2. Adjust the Blue Army target group B-T005-007: reduce one target (from 3 → 2) and adjust the deployment position (from narrow passages → dispersed deployment on both sides) to reduce the density of crossfire and improve the Red Army's mission completion rate; 3. Maintain other Blue Army target parameters unchanged to avoid excessively reducing the intensity of the confrontation and ensure training effectiveness."

[0063] Based on the need to adjust the intensity of the confrontation, a dynamic difficulty control strategy is designed. The strategy includes adjusting the hit rate, reconnaissance range or troop deployment of the blue force, and generating a set of control parameters. This step transforms the adjustment requirements into specific, executable strategies. Combining the control characteristics of the blue team target (supporting three control modes: preset program, intelligent algorithm, and human-in-the-loop, corresponding to easy, medium, and hard difficulty levels), the specific values ​​of the control parameters and their target objects are determined. The specific implementation is as follows: Dynamic difficulty adjustment strategies need to be deeply integrated with the Blue Team's combat formation and control mode to ensure the strategy's feasibility. The strategy design logic and examples are as follows, addressing different adjustment needs: Adjusting the Blue Army's hit rate: The simulated shooting hit rate of the Blue Army targets directly affects the Red Army's survival pressure and shooting opportunities. The adjustment range needs to be set according to the deviation (usually 5%-15% per adjustment to avoid sudden increases or decreases that could lead to training gaps). For example, to address the need to "reduce the hit rate of target B-T003," considering its current difficulty level (medium, driven by intelligent algorithms), its simulated shooting hit rate is reduced from 65% to 55%. Simultaneously, the hit determination threshold is adjusted (from "3 hits determine serious injury to the Red Army" to "4 hits determine serious injury") to ensure the hit rate adjustment matches the actual damage effect. If the target is of easy difficulty (preset program control), the hit rate parameters in the program are directly modified (e.g., changing "6 hits out of 10 shots" to "5 hits out of 10 shots").

[0064] Adjusting the Blue Force's reconnaissance range: The reconnaissance range determines the distance at which the Blue Force targets can detect the Red Force. The larger the range, the more difficult it is for the Red Force to approach covertly. Adjustments need to be made based on battlefield terrain (e.g., a larger range in open areas, a smaller range in complex terrain). For example, if target B-T003 is deployed in a semi-open area (with low walls for cover), and the current reconnaissance range is 500m, and the adjustment requirement is to "reduce it to 400m," then modify the "reconnaissance radius" field in its control parameters from 500m to 400m. Simultaneously adjust the "reconnaissance frequency" (from scanning once every 2 seconds to scanning once every 3 seconds) to further reduce its reconnaissance sensitivity. If the target is at a high difficulty level (human-in-the-loop control), then the reconnaissance range circle needs to be manually dragged in the situational awareness interface at the commander's seat, reducing it from 500m to 400m. The system will automatically generate the corresponding reconnaissance parameters.

[0065] Adjusting the Blue Force's troop deployment: Troop deployment is achieved by increasing or decreasing the number of targets, adjusting target types or positions, and directly affects the Blue Force's defensive density. For example, to address the requirement of "reducing one target in the B-T005-007 target group," the B-T006 target located in the center of the passage (this target is the core of the crossfire) is removed. At the same time, the remaining B-T005 and B-T007 targets are shifted 50m to each side of the passage (from the original spacing of 30m to 80m) to reduce the firepower overlap effect. If a stronger deployment is needed, a new mobile armored target (B-T008) is added and deployed 100m behind the target group as a reserve defensive force to simulate a "Blue Force reinforcement" scenario.

[0066] The generated set of control parameters must be structured in the format of "Control Object ID - Control Type - Original Parameter Value - New Parameter Value - Effective Time - Related Requirements" to ensure that the parameters are traceable and verifiable. For example: "1. Control Target B-T003 (Mobile Rifle Target): Control Type - Hit Rate, Original Parameter 65%, New Parameter 55%; Control Type - Reconnaissance Range, Original Parameter 500m, New Parameter 400m, Effective Time 12:00 (at the start of the next round of combat), Associated Requirement - Improve Red Army Hit Rate; 2. Control Target B-T005-007 (Fixed Rifleman Target Group): Control Type - Troop Deployment, Original Parameter 3 Targets (Location: 112.3380°E / 37.3605°N, 112.3382°E / 37.3607°N, 112.3384°E / 37.3609°N), New Parameter 2 Targets (Keep the first and last two, remove the middle one), Effective Time 12:00, Associated Requirement - Improve Red Army Mission Completion Rate." The parameter set needs to be synchronously stored in the system database to provide a basis for subsequent command issuance.

[0067] The human-in-the-loop control interface allows the director to manually intervene in the Blue Team's actions, or automatically send control parameter sets to the Blue Team's control terminal through the intelligent algorithm-driven mode to generate real-time control commands; This step is the core of implementing the control strategy. Based on the level of difficulty of the confrontation and the needs of the command and control, parameters are sent out under two modes: "human intervention" or "intelligent algorithm-driven," generating executable real-time control commands. The specific implementation is as follows: Human intervention at the control interface (applicable to difficult difficulty levels, requires active decision-making by the controller). The human-in-the-loop control interface needs to be designed with differentiated operating interfaces according to the permissions of different levels of commanders in the Blue Army (company, platoon, squad, vehicle group, individual), ensuring that permissions match the scope of control (e.g., a company commander can adjust the deployment of the entire company, while a platoon leader can only adjust targets within their platoon). For example, for the requirement to "adjust the parameters of target B-T003", a director with "Blue Army company commander" permissions can generate instructions through the following operations: After logging into the control interface, select target B-T003 in the two-dimensional situation map (displaying its current status: hit rate 65%, reconnaissance range 500m, location 112.3385°E / 37.3608°N). Open the "Parameter Adjustment Panel", enter 55% in the "Hit Rate" field and 400m in the "Reconnaissance Range" field. The system will automatically prompt "After adjustment, the threat level of this target to the Red Army by the Blue Army has been reduced from 'High' to 'Medium'". Click "Preview Effect" to load the adjusted virtual situation (the reconnaissance range of B-T003 is reduced from 500m to 400m, and the frequency of simulated shooting hit notifications is reduced). After confirming that everything is correct, click "Issue Command". The system generates standardized manual intervention instructions, including "Instruction ID: CI-001, Intervention type: parameter adjustment, Target ID: B-T003, New parameters: hit rate 55%, reconnaissance range 400m, Issuance time: 11:58:30, Operator: Blue Army Company Commander (ID: BL-001), Execution requirement: effective in the next round of confrontation (12:00)", and the instructions are pushed to the control terminal of B-T003 simultaneously.

[0068] For interventions involving multiple targets, such as "adjusting target deployment," the operator can use the "group selection" function to select B-T005-007 in batches, delete B-T006 in the "deployment adjustment" interface, and drag B-T005 and B-T007 to the new position. This generates a combined instruction containing "target addition / reduction + position offset," ensuring that the intervention logic aligns with tactical intent. Figure 1 To.

[0069] Intelligent algorithm-driven mode (suitable for easy / medium difficulty levels, no manual intervention required) The intelligent algorithm-driven mode automatically reads the control parameter set and combines it with the Blue Force's autonomous engagement rule model (such as the "strong enemy" tactical rule, prioritizing attacks on high-threat targets of the Red Force) to generate standardized single-unit / group combat control commands, eliminating the need for manual operation by the director. For example, the algorithm flow for "B-T003 target parameter adjustment" is as follows: The control parameter set (B-T003: hit rate 55%, reconnaissance range 400m) is read from the database. Combined with the simulated equipment type of the target (motorized gun target) and the combat mission (flank defense), the "Blue Army Single Equipment Parameter Adaptation Algorithm" is invoked. The algorithm automatically converts the parameters into target-recognizable control logic: "55% hit rate → 5-6 hits out of every 10 simulated shots; 400m reconnaissance range → simulated shooting is only triggered on Red Army targets within 400m", and at the same time generates a parameter verification report (such as "55% hit rate is within the reasonable range of 20%-80% for this type of target, and the 400m reconnaissance range meets the tactical reconnaissance radius requirements for mobile gun targets"). According to a unified information transmission protocol (compatible with the Blue Team control terminal), a structured control command is generated: "Command ID: AI-001, Command Type: Parameter Update, Target ID: B-T003, Parameter List: hit_rate=55%, detect_range=400m, Effective Time: 12:00, Generation Algorithm: Blue Team Single-Unit Parameter Adaptation V1.0". The command is sent to the B-T003 control terminal via industrial Ethernet (transmission delay <100ms). After receiving the command, the terminal returns an acknowledgment signal of "Command received, pending activation".

[0070] Regardless of the mode, the generated real-time control commands must include "unique identifier, control content, effective time, execution target, and feedback requirements" to ensure that the control terminal can accurately parse and execute them.

[0071] It executes real-time control commands and monitors the training process, iteratively optimizes control parameters based on real-time feedback data, and achieves closed-loop control of the training process.

[0072] This step is the core implementation link of the closed-loop feedback. By executing instructions, monitoring results, and iteratively optimizing, the gap between the training effect and the target is continuously narrowed, ensuring that the training always progresses around the preset goal. The specific implementation is as follows: Instruction execution and process monitoring Real-time control commands are automatically triggered at a specified effective time (e.g., when the next round of confrontation begins at 12:00). After receiving the command, the Blue Team's control terminal immediately updates the target's operating parameters (e.g., the hit rate module of B-T003 switches from 65% to 55%, and the scanning range of the reconnaissance module is reduced from 500m to 400m). At the same time, it uploads "Command execution status: effective, current parameters: hit rate 55%, reconnaissance range 400m" to the monitoring system.

[0073] The monitoring system needs to collect real-time training data every 10 seconds, focusing on tracking and adjusting performance indicators related to requirements (such as Red Army hit rate, survival rate, and mission completion progress). 12:00-12:30 (first round of confrontation after adjustment), monitoring data shows that: the number of shots fired by the Red Army at B-T003 increased from 25 times / 30 minutes before the adjustment to 32 times (due to reduced interference from the Blue Army's shooting), the number of effective hits increased from 8 times to 12 times, and the hit rate increased from 32% to 37.5% (still lower than 40% of the target). The time it took for the Red Army to advance to the target group area B-T005-007 was reduced from 20 minutes to 15 minutes, and the mission completion rate increased from 75% to 78% (close to 80% of the target). The overall defensive effectiveness of the Blue Team decreased from 90% to 86% (still meeting the target of ≥85%), without excessively weakening the Blue Team's defensive capabilities.

[0074] Monitoring data should be recorded in the format of "timestamp-indicator name-real-time value-target value-deviation" to form a real-time feedback dataset, providing a basis for subsequent iterative optimization.

[0075] Iterative optimization of control parameters Based on real-time feedback data, determine whether the current adjustment parameters meet the training objective. If not, perform a second adjustment. The iteration process must follow the principle of "small adjustments, multiple rounds" (to avoid excessively large adjustments in a single instance that could cause metric fluctuations). For example: After the first round of adjustments, the Red Army's hit rate was 37.5%, which was still 2.5% lower than the target of 40%. Analysis of the feedback data revealed that although B-T003 had been adjusted, the hit rate of the Blue Army's other machine gun target (B-T004) was still 60%, and the interference with the Red Army's flanks and rear was still quite obvious. Generate secondary control requirements: "Fine-tune the hit rate of B-T004 target from 60% to 55% to further reduce interference from the Red Army's flanks and rear." Repeat steps two to three to generate a new set of control parameters and control commands. 12:30-13:00 (After the second adjustment, the monitoring data shows that the Red Army's hit rate has increased to 41% (meeting the standard), the survival rate has increased to 72% (meeting the standard), the mission completion rate has increased to 81% (meeting the standard), and the Blue Army's defense effectiveness is 85% (meeting the standard). At this time, it is determined that "the current parameters meet the training objectives" and the iteration stops. If the standards are still not met, the "feedback-adjustment-monitoring" cycle of optimization continues until the indicators meet the requirements.

[0076] Ultimately, through multiple rounds of closed-loop control, the training effect gradually approached the preset goal, forming a complete closed loop of "evaluation-feedback-adjustment-re-evaluation". This ensured that the red-blue confrontation training could maintain sufficient intensity while allowing the red army to improve its tactical capabilities through adjustments, meeting the core requirements of battalion and company-level live-fire training.

[0077] Another embodiment of the present invention provides an intelligent referee decision-making method in red-blue team training, see [link to relevant documentation]. Figure 2 The method may include: S201 acquires real-time information on the troop status and combat events of both the red and blue forces. It collects location information, equipment status and ammunition consumption data through sensor arrays deployed on live-fire equipment and target terminals, and generates real-time combat situation data by combining digital twin environment mapping. S202 drives the Blue Force target action and calculates the strike effect based on real-time combat situation data. By integrating ballistic models, damage models and autonomous combat rules, it simulates the Blue Force fire strike process and outputs the hit location and damage level determination results. S203 performs damage statistics and capability assessment based on damage level and engagement events, generates combat effectiveness analysis reports by combining preset assessment indicator system, and supports time-series-based two / three-dimensional situational replay and key event reproduction. S204 dynamically adjusts the intensity of the confrontation based on the evaluation results and training objectives, and realizes real-time intervention in the Blue Army's action strategy and closed-loop control of the training process through the human-in-the-loop control interface and intelligent algorithm-driven mode.

[0078] This invention also provides a storage medium storing a computer program, wherein the computer program is configured to execute the steps in any of the above method embodiments when running.

[0079] This invention also provides an electronic device, including a memory and a processor, wherein the memory stores a computer program, and the processor is configured to run the computer program to perform the steps in any of the above method embodiments.

[0080] Specifically, the aforementioned electronic device may further include a transmission device and an input / output device, wherein the transmission device is connected to the aforementioned processor, and the input / output device is connected to the aforementioned processor.

[0081] The above description, based on the embodiments shown in the figures, details the structure, features, and effects of the present invention. The above description is only a preferred embodiment of the present invention, but the present invention is not limited to the scope of implementation shown in the figures. Any changes made in accordance with the concept of the present invention, or equivalent embodiments modified to have equivalent changes, that do not exceed the spirit covered by the specification and figures, should be within the protection scope of the present invention.

Claims

1. An intelligent referee decision-making system for red-blue team training, characterized in that, The system includes: The multi-source battlefield data acquisition module is used to acquire the real-time troop status and combat events of both the red and blue sides. It collects location information, equipment status and ammunition consumption data through sensor arrays deployed on live-fire equipment and target terminals, and generates real-time combat situation data by combining digital twin environment mapping. The intelligent behavior simulation and adversarial adjudication module is used to drive the Blue Force target actions and calculate the strike effect based on real-time combat situation data. By integrating ballistic models, damage models and autonomous combat rules, it simulates the Blue Force fire strike process and outputs the hit location and damage level judgment results. The multi-dimensional battlefield assessment and debriefing analysis module is used to conduct battle damage statistics and capability assessment based on damage level and combat events. It generates combat effectiveness analysis reports by combining preset assessment indicator systems and supports time-series-based two / three-dimensional situation replay and key event reproduction. The closed-loop feedback and guidance intervention module is used to dynamically adjust the intensity of the confrontation based on the evaluation results and training objectives. It realizes real-time intervention in the blue team's action strategy and closed-loop control of the training process through the human-in-the-loop control interface and intelligent algorithm-driven mode.

2. The system according to claim 1, characterized in that, The multi-source battlefield data acquisition module is specifically used for: By deploying multi-source sensor arrays on Red Army individual soldier equipment and Blue Army target terminals, raw data including position coordinates, attitude orientation, equipment operating status and ammunition consumption are collected in real time to generate raw sensor data streams. The raw sensor data stream is preprocessed and fused. The Kalman filter algorithm is used to eliminate noise and compensate for data delay, generating calibrated multi-source state data. The calibrated multi-source state data is input into the digital twin environment, and the red and blue forces in the physical world are mapped one-to-one with the virtual model through the entity mapping algorithm to generate a digital twin entity state mapping table. Based on a digital twin entity state mapping table, and combined with real-time combat event triggers to detect fire strikes and movement behavior, the virtual battlefield situation is dynamically updated, ultimately generating real-time combat situation data.

3. The system according to claim 2, characterized in that, The intelligent behavior simulation and adversarial adjudication module is specifically used for: Analyze the blue force's troop positions, red force's target distribution, and environmental information from real-time combat situation data, input them into the autonomous combat rule model to make behavioral decisions, and generate blue force target action control instructions; According to the Blue Army's target action control instructions, the target is driven to perform concealment, maneuvering or attack actions, while the ballistic model is integrated to calculate the flight trajectory of digital munitions and generate simulated ballistic data. The collision rendezvous is calculated based on simulated ballistic data, and the collision point is detected by combining terrain occlusion and target attitude. The hit location and hit timestamp are output. The damage model is invoked based on the hit location, and the damage effect is calculated by combining the ammunition type and target protection data. Finally, the damage level determination result is output.

4. The system according to claim 3, characterized in that, The multi-dimensional battlefield assessment and debriefing analysis module is specifically used for: Collect damage level assessment results and combat event records, classify and aggregate data according to the red and blue force organizational structure, and generate a battle damage statistics set and an ammunition consumption summary table. Input the battle damage statistics into the preset evaluation index system, calculate the combat effectiveness index score, and the combat effectiveness index includes at least the survival rate, hit rate and mission completion rate to generate preliminary evaluation results; Based on the preliminary assessment results and time-series combat data, an operational effectiveness analysis report is automatically generated, which includes chart-based comparisons of combat damage and analysis of key events. By using a time-series database to replay two- or three-dimensional situations, and by using event triggers to reproduce key combat scenarios, an interactive debriefing and analysis interface is finally output.

5. The system according to claim 4, characterized in that, The closed-loop feedback and guidance intervention module is specifically used for: Analyze the evaluation results in the combat effectiveness analysis report, compare them with the pre-set training target identification deviation, and generate the need to adjust the intensity of the confrontation. Based on the need to adjust the intensity of the confrontation, a dynamic difficulty control strategy is designed. The strategy includes adjusting the hit rate, reconnaissance range or troop deployment of the blue force, and generating a set of control parameters. The human-in-the-loop control interface allows the director to manually intervene in the Blue Team's actions, or automatically send control parameter sets to the Blue Team's control terminal through the intelligent algorithm-driven mode to generate real-time control commands; It executes real-time control commands and monitors the training process, iteratively optimizes control parameters based on real-time feedback data, and achieves closed-loop control of the training process.

6. A smart referee decision-making method in red-blue team training, characterized in that, The method includes: Real-time acquisition of the troop status and combat events of both the red and blue forces; collection of location information, equipment status and ammunition consumption data by sensor arrays deployed on live-fire equipment and target terminals; and generation of real-time combat situation data by combining digital twin environment mapping. Driven by real-time combat situation data, the system simulates the Blue Force's firepower strike process and calculates the strike effect by integrating ballistic models, damage models and autonomous combat rules, and outputs the hit location and damage level determination results. Based on damage levels and combat events, the system performs combat damage statistics and capability assessments, generates combat effectiveness analysis reports by combining preset assessment indicator systems, and supports time-series-based two / three-dimensional situational awareness replay and key event reproduction. The intensity of the confrontation is dynamically adjusted based on the evaluation results and training objectives. Real-time intervention in the Blue Army's action strategy and closed-loop control of the training process are achieved through the human-in-the-loop control interface and intelligent algorithm-driven mode.

7. The method according to claim 6, characterized in that, The real-time acquisition of the troop status and combat events of both the red and blue forces is achieved by collecting location information, equipment status, and ammunition consumption data through sensor arrays deployed on live-fire equipment and target terminals, and combining this data with a digital twin environment mapping to generate real-time combat situation data, including: By deploying multi-source sensor arrays on Red Army individual soldier equipment and Blue Army target terminals, raw data including position coordinates, attitude orientation, equipment operating status and ammunition consumption are collected in real time to generate raw sensor data streams. The raw sensor data stream is preprocessed and fused. The Kalman filter algorithm is used to eliminate noise and compensate for data delay, generating calibrated multi-source state data. The calibrated multi-source state data is input into the digital twin environment, and the red and blue forces in the physical world are mapped one-to-one with the virtual model through the entity mapping algorithm to generate a digital twin entity state mapping table. Based on a digital twin entity state mapping table, and combined with real-time combat event triggers to detect fire strikes and movement behavior, the virtual battlefield situation is dynamically updated, ultimately generating real-time combat situation data.

8. The method according to claim 7, characterized in that, The method, which drives the Blue Force's target actions based on real-time combat situation data and calculates the strike effect, simulates the Blue Force's firepower strike process by integrating ballistic models, damage models, and autonomous combat rules, and outputs the hit location and damage level determination results, including: Analyze the blue force's troop positions, red force's target distribution, and environmental information from real-time combat situation data, input them into the autonomous combat rule model to make behavioral decisions, and generate blue force target action control instructions; According to the Blue Army's target action control instructions, the target is driven to perform concealment, maneuvering or attack actions, while the ballistic model is integrated to calculate the flight trajectory of digital munitions and generate simulated ballistic data. The collision rendezvous is calculated based on simulated ballistic data, and the collision point is detected by combining terrain occlusion and target attitude. The hit location and hit timestamp are output. The damage model is invoked based on the hit location, and the damage effect is calculated by combining the ammunition type and target protection data. Finally, the damage level determination result is output.

9. A storage medium, characterized in that, The storage medium stores a computer program, wherein the computer program is configured to execute the method of any one of claims 6-8 when it is run.

10. An electronic device comprising a memory and a processor, characterized in that, The memory stores a computer program, and the processor is configured to run the computer program to perform the method of any one of claims 6-8.

Images