Single-soldier shooting dynamic feedback analysis method and system based on AI simulation

Abstract

The invention relates to the technical field of intelligent shooting training, and discloses an individual-soldier shooting dynamic feedback analysis method and system based on AI simulation. The method comprises the following steps: constructing a virtual shooting environment model to obtain a physiological adjustment coefficient and a weapon offset coefficient; performing trajectory dynamic prediction; setting a dynamic feedback adjustment strategy to optimize shooting action control; and executing multi-source data fusion and feedback circulation to update the trajectory prediction model. The system comprises a virtual shooting environment construction module, a trajectory dynamic prediction module, a dynamic feedback adjustment strategy execution module and a multi-source data fusion and feedback circulation module. Through multi-dimensional data acquisition and analysis, real-time feedback adjustment and continuous model optimization, the single-soldier shooting training effect is improved, the actual combat ability of soldiers is enhanced, meanwhile, the training cost is reduced, and the training safety is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent shooting training, and in particular to a method and system for analyzing dynamic feedback of individual shooting based on AI simulation. Background Art

[0002] In modern warfare, individual shooting ability is a key component of a soldier's combat effectiveness. Traditional individual shooting training methods have many limitations and cannot meet the high requirements of soldiers' shooting skills in today's complex and changing war environment.

[0003] From the perspective of the training environment, traditional training mostly relies on real shooting ranges, which have limited field conditions and cannot truly simulate a variety of actual combat scenarios. For example, in an urban combat environment, factors such as building obstructions, complex terrain and different lighting conditions are difficult to fully reproduce in conventional shooting range training. Soldiers who train in such a single environment will find it difficult to adapt to the complex environment when entering the real battlefield, which will affect their shooting accuracy and combat effectiveness.

[0004] In terms of data collection and analysis, traditional training methods are relatively backward. In the past, they mainly relied on manual observation and simple measurement tools to record shooting results, such as the location of the bullet impact point. It was impossible to comprehensively and accurately collect other key data during the shooting process, such as changes in the soldier's physiological state, real-time parameters of weapons, and the dynamic impact of environmental factors. This leads to a lack of depth and breadth in training analysis, and it is impossible to explore the factors that affect shooting results from multiple dimensions, making it difficult to provide soldiers with personalized training suggestions.

[0005] In addition, there are serious deficiencies in the feedback mechanism in traditional training. After training, soldiers can often only get simple performance feedback, such as the number of hits, etc. There is a lack of timely and accurate feedback information on specific problems with their own shooting movements, such as weapon shaking caused by gun holding posture and the impact of breathing rhythm on shooting stability. Soldiers find it difficult to understand their subtle mistakes during shooting, and are unable to adjust training methods and improve movements in a timely manner, resulting in slow improvement in training results and prolonged training cycles.

[0006] With the rapid development of artificial intelligence technology, technologies such as deep learning and big data analysis have achieved remarkable results in many fields. However, in the field of individual shooting training, the application of these advanced technologies is still in its infancy and has not yet formed a mature and complete system. At present, there is a lack of systems and methods on the market that can deeply integrate AI technology with individual shooting training to achieve all-round monitoring, real-time analysis and dynamic feedback of the shooting process. Therefore, the development of a dynamic feedback analysis method and system for individual shooting based on AI simulation has important practical significance, which can effectively make up for the shortcomings of traditional training methods and improve the level of individual shooting training. Summary of the invention

[0007] The object of the present invention is to provide a method and system for dynamic feedback analysis of individual soldier shooting based on AI simulation to solve the problems raised in the above-mentioned background technology.

[0008] To achieve the above object, the present invention provides the following technical solution: A method for dynamic feedback analysis of individual soldier shooting based on AI simulation, the method comprising:

[0009] Constructing a virtual shooting environment model, including creating high-precision simulation models of the physiological state of a soldier, weapon parameters, and environmental factors; the virtual shooting environment model integrates real-time physiological sensor data and weapon state data for obtaining physiological adjustment coefficients and weapon offset coefficients;

[0010] Performing ballistic dynamic prediction, including constructing a ballistic prediction model, and obtaining a ballistic prediction result by inputting multi-dimensional shooting data into the ballistic prediction model; the multi-dimensional shooting data includes shooting angle, weapon recoil, environmental wind speed, air humidity, and light intensity;

[0011] Setting a dynamic feedback adjustment strategy, including implementing the dynamic feedback adjustment strategy according to the ballistic prediction result output by the ballistic prediction model; the dynamic feedback adjustment strategy includes real-time correction of training parameters for optimizing shooting action control;

[0012] Performing multi-source data fusion and feedback loop, including periodically collecting actual bullet impact point data and soldier movement data;

[0013] Updating the ballistic prediction model according to the actual bullet impact point data;

[0014] The calculation formula for the weapon offset coefficient is:

[0015] W(t) = α·R(t) + β·∫0 t F(τ)dτ

[0016] where W(t) is the weapon offset coefficient at time t, α is the recoil influence weight, R(t) is the instantaneous recoil value, β is the historical recoil accumulation weight, and F(τ) is the recoil intensity at time τ.

[0017] Preferably, the construction of the virtual shooting environment model includes: simulating a three-dimensional shooting scene using the Unity engine; simulating weapon mechanical vibration and recoil effects through ANSYS; constructing a physiological state prediction model using TensorFlow and inputting heart rate, respiratory rate, and electromyogram signal data.

[0018] Preferably, the acquisition of the multi-dimensional shooting data includes: obtaining the target distance through a laser rangefinder; collecting the weapon attitude angular velocity through an inertial measurement unit; obtaining the wind speed, humidity, and light intensity in real time through an environmental sensor; and measuring the weapon recoil data through a piezoelectric sensor.

[0019] Preferably, the formula for the physiological adjustment coefficient is:

[0020]

[0021] where P(t) is the physiological adjustment coefficient at time t, γ i is the adjustment weight of the i-th physiological parameter, S i (t) is the real-time value of the i-th physiological parameter, and S i,ref is the reference value of the i-th physiological parameter.

[0022] Preferably, the training process of the ballistic prediction model includes:

[0023] Data collection: Collect 10,000 groups of shooting data from the simulated shooting system, including shooting angle, recoil, environmental parameters, and actual impact point deviation;

[0024] Feature engineering: Normalize the shooting data, convert the angle data into a three-dimensional vector in the spherical coordinate system, and extract the frequency domain features of the weapon attitude;

[0025] Model construction: Use a temporal convolutional network to process time series data and focus on the key shooting stages through an attention mechanism layer;

[0026] Verification and optimization: Divide the data set into a training set and a test set according to 8:2, and use the cross-entropy loss function for model training.

[0027] Preferably, the dynamic feedback adjustment strategy includes: designing a feedback controller based on the reinforcement learning algorithm and generating real-time correction instructions according to the ballistic prediction results; the calculation formula for the correction instructions is:

[0028]

[0029] where C(t) is the correction instruction, K p , K i , K d are the proportional, integral, and differential gains respectively, e(t) is the current impact point deviation, E(t) is the future deviation output by the ballistic prediction model, and K f is the prediction gain coefficient.

[0030] Preferably, the multi-source data fusion and feedback loop includes: performing Kalman filter fusion on the actual impact point data and the prediction results of the virtual model to generate a calibrated ballistic trajectory; collecting soldier limb movement data through an optical motion capture system for optimizing the weight allocation of the physiological adjustment coefficients.

[0031] Preferably, the prediction process of the ballistic prediction model includes:

[0032] Receiving real-time shooting angle, weapon attitude, environmental parameters, and recoil data;

[0033] Extracting time series features through TCN and assigning importance to each time step through attention weights;

[0034] Performing tensor splicing on the spatio-temporal features, physiological adjustment coefficients, and weapon offset coefficients;

[0035] Predicting future impact point coordinates and deviation values through a fully connected network.

[0036] Preferably, the model update uses an online incremental learning algorithm to dynamically adjust the network weights of TCN according to the latest collected impact point data, and adopts a sliding window mechanism to eliminate noise samples in historical data.

[0037] Preferably, the present invention further includes a single-soldier shooting dynamic feedback analysis system based on AI simulation, and the system includes:

[0038] Virtual shooting environment construction module: used to construct a virtual shooting environment model, specifically create a high-precision simulation model of the soldier's physiological state, weapon parameters, and environmental factors, and integrate real-time physiological sensor data and weapon state data to obtain physiological adjustment coefficients and weapon offset coefficients;

[0039] Ballistic dynamic prediction module: construct a ballistic prediction model, receive multi-dimensional shooting data including shooting angle, weapon recoil, environmental wind speed, air humidity, and light intensity, and output ballistic prediction results;

[0040] Dynamic feedback adjustment strategy execution module: according to the ballistic prediction results output by the ballistic prediction module, implement a dynamic feedback adjustment strategy, perform real-time correction on training parameters, and optimize shooting action control;

[0041] Multi-source data fusion and feedback loop module: periodically collect actual impact point data and soldier movement data, and update the ballistic prediction model according to the actual impact point data;

[0042] Among them, the calculation formula of the weapon offset coefficient is:

[0043] W(t) = α·R(t) + β·∫0 t F(τ)dτ

[0044] Among them, W(t) is the weapon offset coefficient at time t, α is the recoil influence weight, R(t) is the instantaneous recoil value, β is the historical recoil accumulation weight, and F(τ) is the recoil intensity at time τ.

[0045] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0046] The present invention constructs a virtual shooting environment model and creates a high-precision simulation model of the soldier's physiological state, weapon parameters, and environmental factors. The Unity engine is used to simulate a three-dimensional shooting scene, ANSYS is used to simulate weapon mechanical vibration and recoil effects, and TensorFlow is used to construct a physiological state prediction model, which can highly reproduce the real battlefield environment. When soldiers train in such a virtual environment, they can adapt to various complex combat scenarios in advance, such as shooting situations under different terrains, weather conditions, and lighting conditions. Compared with the simplicity of traditional training environments, the diversity of this virtual environment enables soldiers to respond more quickly and accurately when facing actual combat, greatly enhancing their actual combat adaptability.

[0047] The invention covers multi-dimensional shooting data collection. Through a laser rangefinder, an inertial measurement unit, environmental sensors, piezoelectric sensors, etc., data such as the target distance, weapon attitude angular velocity, wind speed, humidity, light intensity, and weapon recoil are comprehensively obtained. At the same time, the physiological data of soldiers, such as heart rate, breathing frequency, and electromyogram signals, are collected to calculate the physiological adjustment coefficient. These rich data provide a basis for in-depth analysis of the shooting process. By analyzing the data, factors affecting shooting accuracy can be accurately located, a personalized training plan can be formulated for soldiers, and more targeted training guidance can be provided, effectively improving the training effect.

[0048] Implement a dynamic feedback adjustment strategy based on the results of the ballistic prediction model. A feedback controller is designed through a reinforcement learning algorithm to generate real-time correction instructions. This instruction is calculated based on the current impact point deviation and the predicted future deviation of the ballistic trajectory, and can timely correct the soldier's shooting action. During the soldier's shooting process, the system continuously adjusts the training parameters according to real-time data to optimize the control of shooting actions. Different from the lagging feedback method in traditional training, this real-time dynamic feedback enables soldiers to immediately understand their action problems and make adjustments, continuously strengthening correct shooting actions and reducing the repetition of incorrect actions, significantly improving the speed of shooting skill improvement.

[0049] The ballistic prediction model is updated using an online incremental learning algorithm and a sliding window mechanism. As new impact point data is continuously collected, the model can dynamically adjust the network weights while eliminating noise samples in historical data. This enables the model to always maintain its adaptability to the latest shooting data and continuously improve the prediction accuracy. Whether it is a minor change in weapon performance or an adjustment in the shooting habits of soldiers, the model can respond in a timely manner, providing accurate ballistic predictions and feedback for training to ensure that the accuracy and timeliness of training always remain at a high level.

[0050] The construction of a virtual shooting environment reduces the dependence on a large number of live firing range trainings, lowering training costs such as venue rental and ammunition consumption. At the same time, it avoids potential safety accidents in live ammunition training, such as weapon malfunctions and accidental shootings, improving the safety of training. For shooting training scenarios in some dangerous environments, such as high-radiation areas or chemically polluted areas, virtual training provides a safe and feasible training method, achieving the training objectives while ensuring the safety of soldiers' lives. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] Figure 1 is the working principle diagram of the single-soldier shooting dynamic feedback analysis method based on AI simulation described in the present invention;

[0052] Figure 2 is the construction flow chart of the virtual shooting environment model;

[0053] Figure 3 is the working flow chart of multi-dimensional shooting data collection;

[0054] Figure 4 is the working flow chart of the execution of the dynamic feedback adjustment strategy. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0055] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without making creative efforts shall fall within the protection scope of the present invention.

[0056] Please refer to Figures 1-4 , the present invention provides a technical solution: a single-soldier shooting dynamic feedback analysis method based on AI simulation, the method includes:

[0057] Construct a virtual shooting environment model: Create a high-precision simulation model of the soldier's physiological state, weapon parameters, and environmental factors, integrate real-time physiological sensor data and weapon state data, and obtain physiological adjustment coefficients and weapon offset coefficients.

[0058] Conduct ballistic dynamic prediction: Construct a ballistic prediction model, input multi-dimensional shooting data such as shooting angle, weapon recoil, environmental wind speed, air humidity, and light intensity, and obtain the ballistic prediction result.

[0059] Set a dynamic feedback adjustment strategy: Implement a dynamic feedback adjustment strategy based on the ballistic prediction result, perform real-time correction on the training parameters, and optimize the shooting action control.

[0060] Execute multi-source data fusion and feedback loop: Periodically collect actual impact point data and soldier movement data, and update the ballistic prediction model according to the actual impact point data. The calculation formula for the weapon offset coefficient is:

[0061] W(t) = α·R(t) + β·∫0 t F(τ)dτ

[0062] Where, W(t) is the weapon offset coefficient at time t, α is the influence weight of recoil, R(t) is the instantaneous recoil value, β is the cumulative weight of historical recoil, and F(τ) is the recoil intensity at time τ.

[0063] The present invention will be further described below in conjunction with Examples 1 to 5:

[0064] Example 1:

[0065] This example mainly elaborates on the specific method for constructing a virtual shooting environment model, and its function is to realize a more realistic and accurate virtual shooting environment simulation through a variety of professional tools and technologies.

[0066] Use the Unity engine to simulate a three-dimensional shooting scene. The Unity engine has powerful graphics rendering capabilities and rich plugin resources, and can quickly build a realistic shooting scene, including elements such as terrain, buildings, and targets. When constructing the scene, highly restore the details of the terrain. For example, different landform features (mountains, plains, jungles, etc.) will affect the shooting line of sight and bullet flight trajectory, and these situations can be simulated through precise modeling. For buildings, finely design their structures and materials, and consider the rebound and penetration effects when bullets hit different materials. The design of the target is based on training requirements, setting different shapes, sizes, and moving speeds to simulate diverse targets in actual combat.

[0067] Simulate the mechanical vibration and recoil effect of weapons through ANSYS. ANSYS is a professional engineering simulation software with significant advantages in weapon simulation. It can model the mechanical structure of weapons in detail and analyze the mechanical properties of weapons during shooting. When simulating the mechanical vibration of weapons, accurately calculate the vibration amplitude, frequency of each component of the weapon at the moment of firing a bullet, and how these vibrations are transmitted to the soldier's hand, affecting the shooting stability. For the recoil effect, simulate the magnitude, direction, and action time of the recoil to provide data support for obtaining accurate weapon offset coefficients.

[0068] Build a physiological state prediction model using TensorFlow, and input heart rate, respiratory rate, and electromyogram signal data. TensorFlow is a widely used deep learning framework. Using it to build a physiological state prediction model can effectively process complex physiological data. Collect a large amount of data on heart rate, respiratory rate, and electromyogram signals of soldiers in different shooting states, preprocess these data to remove noise and outliers. Then, use neural network structures such as recurrent neural network (RNN) or long short-term memory network (LSTM) to train the data. During the training process, continuously adjust the network parameters so that the model can accurately predict the changes in the physiological state of soldiers, and then obtain more accurate physiological adjustment coefficients.

[0069] Example 2:

[0070] This example details the acquisition method of multi-dimensional shooting data, whose important role is to ensure the acquisition of comprehensive and accurate shooting data, providing a reliable basis for ballistic dynamic prediction and subsequent analysis.

[0071] Obtain the target distance through a laser rangefinder. The laser rangefinder measures the distance by using the propagation time of the laser beam, featuring high precision and fast measurement speed. In practical applications, install the laser rangefinder at a suitable position on the weapon to ensure that its measurement line of sight is parallel or has a known angular relationship with the weapon's aiming line. When the soldier aims at the target, the laser rangefinder emits a laser beam towards the target. After the laser beam hits the target and reflects back, the rangefinder calculates the target distance based on the round-trip time of the laser beam. To improve the measurement accuracy, the laser rangefinder can also be equipped with a data processing module to filter and average the data of multiple measurements and remove outliers.

[0072] Collect the weapon attitude angular velocity through an inertial measurement unit. The inertial measurement unit (IMU) usually consists of an accelerometer and a gyroscope, and can measure the acceleration and angular velocity of an object in real time. Install the IMU at key parts of the weapon, such as the gun body or the grip, to ensure that it can accurately sense the attitude changes of the weapon. During shooting, the IMU continuously collects the weapon attitude angular velocity data, which reflects the rotational speed of the weapon in all directions. By analyzing these data, the jitter situation of the weapon when the soldier shoots can be understood, providing important information for evaluating shooting stability.

[0073] Obtain the wind speed, humidity, and light intensity in real time through environmental sensors. The environmental sensors include a wind speed sensor, a humidity sensor, and a light sensor. The wind speed sensor generally uses a cup anemometer or an ultrasonic type, and is installed in an open position of the shooting range to accurately measure the current wind speed magnitude and direction. The humidity sensor is used to measure the humidity in the air, and its measurement principle can be capacitive or resistive. Install it in a suitable position to ensure that the humidity data of the shooting environment can be accurately obtained. The light sensor is used to measure the light intensity. Different light intensities will affect the aiming accuracy of the soldier. By collecting these data, the influence of the light factor can be considered in the ballistic prediction model. The weapon recoil force data is measured by a piezoelectric sensor. The piezoelectric sensor is installed at the part where the weapon contacts the soldier, such as the buttstock or the grip. When the weapon generates recoil force, the piezoelectric sensor is subjected to pressure to generate an electrical signal. Through the processing and conversion of the electrical signal, accurate recoil force data is obtained.

[0074] Example 3:

[0075] This example elaborates on the calculation method of the physiological adjustment coefficient and its application in shooting analysis. Its function is to provide a basis for optimizing shooting actions by quantifying the influence of physiological factors on shooting.

[0076] The formula for the physiological adjustment coefficient is:

[0077]

[0078] where P(t) is the physiological adjustment coefficient at time t, γ i is the adjustment weight of the i-th physiological parameter, S i (t) is the real-time value of the i-th physiological parameter, S i,ref is the reference value of the i-th physiological parameter.

[0079] In practical applications, first determine various physiological parameters and their reference values. For example, the reference value of the heart rate can be set as the average heart rate of a soldier in a calm state, and the reference value of the breathing rate is the middle value within the normal breathing rate range. Through physiological sensors worn on the soldier, such as a heart rate monitor, a breathing rate sensor, and a muscle electrical signal collector, the real-time values of various physiological parameters are obtained in real time.

[0080] According to the importance of the influence of different physiological parameters on shooting, determine the adjustment weight γ. i For example, the change in heart rate has a greater impact on shooting stability, and its adjustment weight can be set relatively high; while certain muscle electrical signals have a more significant impact under specific shooting actions, and the corresponding adjustment weights can also be adjusted according to the actual situation.

[0081] Calculate the physiological adjustment coefficient P(t). When the soldier is in the process of shooting, various physiological parameters change in real time, and the physiological adjustment coefficient calculated through the above formula can reflect the comprehensive influence of the current physiological state on shooting. For example, when the soldier's heart rate increases and breathing becomes rapid, the calculated physiological adjustment coefficient will increase, indicating that the physiological state has a greater negative impact on shooting at this time, and it is necessary to optimize the shooting effect by adjusting the shooting action or performing psychological adjustment. In the ballistic prediction model, the physiological adjustment coefficient is input as an important parameter and participates in the calculation together with other shooting data, enabling the model to more accurately predict the ballistic trajectory.

[0082] Example 4:

[0083] The training process of the ballistic prediction model is as follows:

[0084] Data collection: Collect 10,000 groups of shooting data from the simulated shooting system, including shooting angles, recoil forces, environmental parameters, and actual impact point deviations. These data cover various shooting conditions, such as different shooting distances, different environmental wind speeds and humidities, different weapon recoil force magnitudes, etc., to ensure the diversity and comprehensiveness of the model training data.

[0085] Feature engineering: Normalize the shooting data, convert the angle data into a three-dimensional vector in the spherical coordinate system, and extract the frequency domain features of the weapon posture. Normalization can make the data of different features have the same scale, facilitating model training and comparison. Converting the angle data into a three-dimensional vector in the spherical coordinate system can more intuitively represent the shooting direction and is more convenient to process angle-related information in subsequent model calculations. Extracting the frequency domain features of the weapon posture can analyze the movement state of the weapon from another angle and discover some laws that are not easily detected in the time domain.

[0086] Model construction: The Time Convolutional Network (TCN) is used to process time series data, and the attention mechanism layer is used to focus on the key shooting stages. TCN has powerful time series processing capabilities and can effectively capture the time-dependent relationships in shooting data. The attention mechanism layer can dynamically assign weights according to the importance of the data, enabling the model to pay more attention to the key shooting stages, such as various data changes at the moment of bullet firing.

[0087] Verification and optimization: The dataset is divided into a training set and a test set in a ratio of 8:2, and the cross-entropy loss function is used for model training. During the training process, the model parameters are continuously adjusted to gradually reduce the loss value of the model on the training set. At the same time, the test set is used to verify the model to ensure that the model has good generalization ability and avoid overfitting.

[0088] The prediction process of the ballistic prediction model is as follows:

[0089] Receiving data: Receive real-time shooting angle, weapon attitude, environmental parameters, and recoil data. These data are obtained in real time through the above data acquisition methods and are transmitted to the ballistic prediction model in a timely manner.

[0090] Feature extraction and processing: Extract time series features through TCN and assign the importance of each time step through attention weights. TCN performs convolution operations on the input time series data to extract key features. The attention mechanism assigns corresponding weights according to the importance of data at different time steps, enabling the model to pay more attention to the data that has a greater impact on the prediction results.

[0091] Tensor concatenation: Concatenate spatio-temporal features with physiological adjustment coefficients and weapon offset coefficients. The extracted spatio-temporal features are combined with physiological adjustment coefficients and weapon offset coefficients calculated by other methods to form a comprehensive feature tensor, providing more comprehensive information for subsequent predictions.

[0092] Output of prediction results: Predict the future impact point coordinates and deviation values through a fully connected network. The fully connected network performs complex non-linear transformations based on the input comprehensive feature tensor and finally outputs the predicted future impact point coordinates and deviation values, providing a basis for subsequent dynamic feedback adjustment strategies.

[0093] Example 5:

[0094] The dynamic feedback adjustment strategy includes: Designing a feedback controller based on the reinforcement learning algorithm and generating real-time correction instructions according to the ballistic prediction results. The reinforcement learning algorithm aims to maximize the cumulative reward by allowing the agent to continuously try and learn in the environment. In the present invention, the feedback controller serves as the agent, and based on the ballistic prediction results output by the ballistic prediction model and combined with the current shooting state, real-time correction instructions are generated. The calculation formula for the correction instructions is:

[0095]

[0096] Among them, C(t) is the correction instruction, K p , K i , K d are the proportional, integral, and derivative gains respectively, e(t) is the current impact point deviation, E(t) is the future deviation output by the ballistic prediction model, and K f is the prediction gain coefficient.

[0097] In practical applications, according to different shooting scenarios and training requirements, the proportional, integral, and derivative gains and the prediction gain coefficient are adjusted. For example, in close-range shooting, since the target distance is relatively close and the bullet flight time is short, more attention needs to be paid to the correction of the current impact point deviation. At this time, the proportional gain K p can be appropriately increased; while in long-range shooting, various factors during the bullet flight need to be considered, and the integral gain K i and the prediction gain coefficient K f play a more important role, and their values can be adjusted accordingly.

[0098] The model is updated using an online incremental learning algorithm, which dynamically adjusts the network weights of the TCN according to the latest collected impact point data, and uses a sliding window mechanism to eliminate noise samples in the historical data. The online incremental learning algorithm enables the model to continuously update the network weights in the case of continuously obtaining new data, adapting to changes in the shooting environment and the shooting state of soldiers. When new impact point data is collected, it is input into the model, and the network weights of the TCN are adjusted through the backpropagation algorithm, enabling the model to more accurately predict the ballistic trajectory. The sliding window mechanism sets a fixed-size window in the historical data. As new data continuously enters, the data within the window is continuously updated. During the window update process, the data within the window is analyzed to identify and eliminate noise samples, ensuring the quality of the model training data, thereby improving the accuracy and stability of the model.

[0099] The present invention also includes a single-soldier shooting dynamic feedback analysis system based on AI simulation, and the system includes:

[0100] Virtual shooting environment construction module: used to construct a virtual shooting environment model, specifically create high-precision simulation models of soldiers' physiological states, weapon parameters, and environmental factors, and integrate real-time physiological sensor data and weapon state data to obtain physiological adjustment coefficients and weapon offset coefficients;

[0101] Ballistic dynamic prediction module: constructs a ballistic prediction model, receives multi-dimensional shooting data including shooting angle, weapon recoil, environmental wind speed, air humidity, and light intensity, and outputs a ballistic prediction result;

[0102] Dynamic feedback adjustment strategy execution module: According to the ballistic prediction result output by the ballistic prediction module, implement the dynamic feedback adjustment strategy, perform real-time correction on the training parameters, and optimize the shooting action control;

[0103] Multi-source data fusion and feedback loop module: Periodically collect actual impact point data and soldier movement data, and update the ballistic prediction model according to the actual impact point data;

[0104] Among them, the calculation formula of the weapon offset coefficient is:

[0105] W(t) = α·R(t) + β·∫0 t F(τ)dτ

[0106] Among them, W(t) is the weapon offset coefficient at time t, α is the influence weight of recoil, R(t) is the instantaneous recoil value, β is the cumulative weight of historical recoil, and F(τ) is the recoil intensity at time τ.

[0107] The implementation manner of this system refers to the above-mentioned embodiments and will not be elaborated in the specification.

[0108] It should be noted that in this article, relational terms such as first and second are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variant thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements not only includes those elements, but also includes other elements not expressly listed, or also includes elements inherent to such process, method, article or device.

[0109] Although the embodiments of the present invention have been shown and described, for those of ordinary skill in the art, it can be understood that various changes, modifications, substitutions and variations can be made to these embodiments without departing from the principle and spirit of the present invention. The scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A method for analyzing the dynamic feedback of individual soldier shooting based on AI simulation, characterized in that, Including: Constructing a virtual shooting environment model, including creating high-precision simulation models of soldiers' physiological states, weapon parameters, and environmental factors; the virtual shooting environment model integrates real-time physiological sensor data and weapon state data to obtain physiological adjustment coefficients and weapon offset coefficients; Performing ballistic dynamic prediction, including constructing a ballistic prediction model, and obtaining a ballistic prediction result by inputting multi-dimensional shooting data into the ballistic prediction model; the multi-dimensional shooting data includes shooting angle, weapon recoil, environmental wind speed, air humidity, and light intensity; Setting a dynamic feedback adjustment strategy, including implementing the dynamic feedback adjustment strategy according to the ballistic prediction result output by the ballistic prediction model; the dynamic feedback adjustment strategy includes real-time correction of training parameters for optimizing shooting action control; Executing a multi-source data fusion and feedback loop, including periodically collecting actual impact point data and soldier movement data; Updating the ballistic prediction model according to the actual impact point data; The calculation formula for the weapon offset coefficient is: Where, W(t) is the weapon offset coefficient at time t, α is the recoil influence weight, R(t) is the instantaneous recoil value, β is the historical recoil accumulation weight, and F(τ) is the recoil intensity at time τ.

2. The method for analyzing the dynamic feedback of individual soldier shooting based on AI simulation according to claim 1, wherein The construction of the virtual shooting environment model includes: using the Unity engine to simulate a three-dimensional shooting scene; simulating weapon mechanical vibration and recoil effects through ANSYS; using TensorFlow to construct a physiological state prediction model and inputting heart rate, respiratory rate, and electromyogram signal data.

3. An AI-simulated individual soldier shooting dynamic feedback analysis method according to claim 1, characterized in that The acquisition of the multi-dimensional shooting data includes: obtaining the target distance through a laser rangefinder; collecting the weapon attitude angular velocity through an inertial measurement unit; obtaining wind speed, humidity, and light intensity in real time through an environmental sensor; the weapon recoil data is measured by a piezoelectric sensor.

4. A method for analyzing the dynamic feedback of individual soldier shooting based on AI simulation according to claim 1, characterized in that, The formula for the physiological adjustment coefficient is: where P(t) is the physiological adjustment coefficient at time t, and γ i is the adjustment weight of the i-th physiological parameter, and S i (t) is the real-time value of the i-th physiological parameter, and S i,ref is the reference value of the i-th physiological parameter.

5. A dynamic feedback analysis method for individual soldier shooting based on AI simulation according to claim 1, characterized in that The training process of the ballistic prediction model includes: Data collection: Collect 10,000 groups of shooting data from the simulated shooting system, including shooting angle, recoil, environmental parameters, and actual impact point deviation; Feature engineering: Normalize the shooting data, convert the angle data into a three-dimensional vector in spherical coordinates, and extract the frequency domain features of the weapon attitude; Model construction: Use a temporal convolutional network to process time series data and focus on key shooting stages through an attention mechanism layer; Verification and optimization: Divide the data set into a training set and a test set according to 8:2, and use a cross-entropy loss function for model training.

6. The method for analyzing the dynamic feedback of individual soldier shooting based on AI simulation according to claim 1, wherein The dynamic feedback adjustment strategy includes: Designing a feedback controller based on a reinforcement learning algorithm, and generating a real-time correction instruction according to the ballistic prediction result; the calculation formula for the correction instruction is: Among them, C(t) is the correction instruction, K p , K i , K d are the proportional, integral, and derivative gains respectively, e(t) is the current impact point deviation, E(t) is the future deviation output by the ballistic prediction model, and K f is the prediction gain coefficient.

7. A method for analyzing the dynamic feedback of individual soldier shooting based on AI simulation according to claim 1, characterized in that The multi-source data fusion and feedback loop includes: Performing Kalman filter fusion on the actual impact point data and the prediction result of the virtual model to generate a calibrated ballistic trajectory; collecting soldier limb movement data through an optical motion capture system for optimizing the weight allocation of the physiological adjustment coefficient.

8. A method for analyzing the dynamic feedback of individual soldier shooting based on AI simulation according to claim 5, characterized in that The prediction process of the ballistic prediction model includes: Receiving real-time shooting angle, weapon attitude, environmental parameters, and recoil data; Extract time series features through TCN and assign the importance of each time step through attention weights; Perform tensor splicing on spatio-temporal features, physiological adjustment coefficients, and weapon offset coefficients; Predict the future impact point coordinates and deviation values through a fully connected network.

9. A method for analyzing the dynamic feedback of individual soldier shooting based on AI simulation according to claim 1, characterized in that, The model update adopts an online incremental learning algorithm, dynamically adjusts the network weights of TCN according to the latest collected impact point data, and uses a sliding window mechanism to eliminate noise samples in historical data.

10. A single-soldier shooting dynamic feedback analysis system based on AI simulation, characterized in that, It includes: Virtual shooting environment construction module: used to construct a virtual shooting environment model, specifically create high-precision simulation models of soldiers' physiological states, weapon parameters, and environmental factors, and integrate real-time physiological sensor data and weapon state data to obtain physiological adjustment coefficients and weapon offset coefficients; Ballistic dynamic prediction module: construct a ballistic prediction model, receive multi-dimensional shooting data including shooting angle, weapon recoil, environmental wind speed, air humidity, and light intensity, and output ballistic prediction results; Dynamic feedback adjustment strategy execution module: implement a dynamic feedback adjustment strategy based on the ballistic prediction results output by the ballistic prediction module, perform real-time correction on training parameters, and optimize shooting action control; Multi-source data fusion and feedback loop module: periodically collect actual impact point data and soldier action data, and update the ballistic prediction model according to the actual impact point data; Among them, the calculation formula of the weapon offset coefficient is: Among them, W(t) is the weapon offset coefficient at time t, α is the recoil influence weight, R(t) is the instantaneous recoil value, β is the historical recoil accumulation weight, and F(τ) is the recoil intensity at time τ.

Images