A tactical simulation training method for wheeled assault vehicles in different environments

Abstract

The present invention discloses a tactical simulation training method for wheeled assault vehicles in different environments, comprising the following steps: obtaining a multi-scale environmental parameter vector of the current position of the wheeled assault vehicle; adopting a hybrid digital twin model consisting of a basic performance model based on physical mechanisms and a data-driven residual compensation model for the key subsystems of the vehicle, wherein the model combines the driving behavior vector and the environmental parameter vector to calculate the compensated performance parameters; updating the accumulated loss state variables of the key subsystems in real time according to the performance parameters; feeding back the accumulated loss state variables to the basic performance model to correct its parameters in real time, thereby enabling the vehicle performance to dynamically and realistically decay with the accumulated loss; the present invention can realistically simulate the "non-combat loss" of the vehicle, deeply bind driving behavior with equipment health management, and force the driver to consider equipment loss in tactical decision-making, thereby expanding the dimension of training and improving the authenticity and effectiveness of tactical simulation training.

Description

Technical Field

[0001] The present invention relates to the technical field of military vehicle simulation training, in particular to a tactical simulation training method for wheeled assault vehicles in different environments. Background Art

[0002] Currently, mainstream tactical simulation training systems mostly employ event-triggered physics models to address environmental factors. For example, when a vehicle enters muddy terrain, it simulates the current environmental conditions by increasing driving resistance. This approach treats the vehicle as an idealized model with fixed performance parameters. Its performance, aside from direct combat damage such as enemy fire, rarely degrades naturally with combat time, mileage, or environmental changes.

[0003] This simulation training logic leads to two core flaws, severely misaligning the effectiveness of simulation training with actual combat requirements. First, it ignores the continuous and cumulative wear and tear that environmental changes cause to a vehicle's internal subsystems. In real combat, wheeled assault vehicles experience irreversible tire wear after prolonged traversal of harsh terrain such as mud, desert, or gravel. Air filters can become clogged to varying degrees due to environmental fluctuations, and the suspension suffers cumulative fatigue damage from constant turbulence. These "non-combat wear and tear" gradually and realistically degrade the vehicle's maneuverability, reliability, and overall combat effectiveness. Existing simulation training systems, however, allow drivers to engage in unrestrained, high-intensity maneuvers without risking the consequences of vehicle performance degradation. Second, training lacks the dimension of managing equipment "health" and tactical maintenance decisions. Because vehicles are designed as ideal models with virtually no fatigue or wear and tear, drivers' focus is easily limited to the "search-fire" tactical cycle. This fails to cultivate drivers' ability to perceive vehicle status, predict potential failures, and make equipment maintenance and tactical decisions in real combat.

[0004] Therefore, there is an urgent need for a simulation training method that can take into account the cumulative loss of simulated vehicles and integrate tactical maintenance decisions to make up for the shortcomings of the current simulation training system. Summary of the Invention

[0005] The purpose of this section is to summarize some aspects of the embodiments of the present invention and briefly introduce some preferred embodiments. Some simplifications or omissions may be made in this section and the abstract and title of this application to avoid obscuring the purpose of this section, the abstract and the title of the invention, and such simplifications or omissions should not be used to limit the scope of the present invention.

[0006] In view of the above existing problems, the present invention is proposed. Therefore, the present invention provides a wheeled assault vehicle tactical simulation training method in different environments to solve the problems raised in the background technology.

[0007] To solve the above technical problems, the present invention provides the following technical solutions: a tactical simulation training method for wheeled assault vehicles in different environments, comprising:

[0008] Obtain the multi-scale environmental parameter vector of the current position of the wheeled assault vehicle in simulation training;

[0009] Calculating compensated performance parameters of at least one key subsystem of the wheeled assault vehicle using a hybrid digital twin model, wherein the hybrid digital twin model is composed of a physical mechanism-based basic performance model and a data-driven residual compensation model;

[0010] The basic performance model calculates the basic performance parameters of the key subsystem according to the driving behavior vector;

[0011] The residual compensation model calculates a residual compensation value for the basic performance parameter based on the multi-scale environmental parameter vector and the driving behavior vector;

[0012] Calculating the accumulated loss state variable of the key subsystem based on the compensated performance parameter of the key subsystem;

[0013] The performance status of the key subsystems in the simulation training is dynamically updated based on the feedback of the accumulated loss state variable to the basic performance model.

[0014] As a preferred solution of the wheeled assault vehicle tactical simulation training method for different environments described in the present invention, the step of obtaining a multi-scale environmental parameter vector includes:

[0015] Load preset digital map data containing elevation, surface material and weather information;

[0016] Acquiring the geographical location coordinates of the wheeled assault vehicle in real time;

[0017] According to the geographic location coordinates, parameters including at least a road surface roughness index, a road surface adhesion coefficient, a road surface abrasion coefficient, and a filter clogging coefficient are parsed from the digital map data to form the multi-scale environmental parameter vector.

[0018] As a preferred solution of the wheeled assault vehicle tactical simulation training method for different environments described in the present invention, wherein: the residual compensation model is a machine learning model obtained by offline training of the gradient boosting decision tree algorithm;

[0019] The input of the model is a feature vector formed by splicing the multi-scale environmental parameter vector and the driving behavior vector, and the output is the residual compensation value.

[0020] As a preferred solution of the wheeled assault vehicle tactical simulation training method for different environments described in the present invention, the offline training method of the gradient boosting decision tree algorithm includes:

[0021] Collect synchronized data points containing real-world environment data, real-world driving data, and real-world subsystem performance data;

[0022] For each synchronous data point, inputting the actual driving data into the basic performance model to obtain a theoretical prediction value;

[0023] Calculating the difference between the actual subsystem performance data and the theoretical prediction value as a training label;

[0024] splicing the real environment data and the real driving data as training features;

[0025] The gradient boosting decision tree model is trained using the generated training dataset consisting of {feature, label} pairs.

[0026] As a preferred solution of the wheeled assault vehicle tactical simulation training method for different environments described in the present invention, the key subsystems include at least a tire subsystem, a suspension subsystem and a powertrain subsystem.

[0027] As a preferred solution of the wheeled assault vehicle tactical simulation training method for different environments described in the present invention, the step of calculating the cumulative loss state variable of the key subsystem includes:

[0028] For the tire subsystem, calculating the increment of tire wear depth according to the dynamic vertical load, longitudinal slip rate, sideslip angle and road abrasion coefficient of the tire;

[0029] For the suspension subsystem, the fatigue damage degree of key suspension components is calculated based on the dynamic travel of the suspension using the rain flow counting method and the preset SN curve;

[0030] For the powertrain subsystem, the blockage mass increment of the filter is calculated according to the instantaneous intake air flow of the engine and the concentration of particulate matter in the ambient air.

[0031] As a preferred solution of the wheeled assault vehicle tactical simulation training method for different environments described in the present invention, it also includes:

[0032] Presetting at least one early warning threshold value of the cumulative loss state variable;

[0033] When any accumulated loss state variable reaches its corresponding warning threshold, graded warning information is provided to the driver.

[0034] As a preferred solution of the wheeled assault vehicle tactical simulation training method for different environments described in the present invention, it also includes:

[0035] According to the current cumulative loss state of the wheeled assault vehicle, a tactical task with constraints that matches the current vehicle performance state is dynamically generated.

[0036] As a preferred solution of the wheeled assault vehicle tactical simulation training method for different environments described in the present invention, it also includes:

[0037] Based on the current cumulative loss status of the wheeled assault vehicle, mission time limit and available maintenance resources, at least one time-conflict, resource-conflict or space-conflict tactical decision-making game scenario is constructed to force the driver to make a trade-off between equipment maintenance and mission execution.

[0038] As a preferred solution of the wheeled assault vehicle tactical simulation training method for different environments described in the present invention, it also includes:

[0039] Calculate and update a comprehensive equipment preservation score in real time based on the real-time cumulative loss value of each key subsystem and its preset impact weight on the combat effectiveness of the entire vehicle;

[0040] After the simulation training is completed, an interactive review interface is generated that includes the vehicle's driving trajectory, equipment preservation score curve, and loss curves of key vehicle components for training evaluation and teaching analysis.

[0041] Compared with the prior art, the invention has the following beneficial effects:

[0042] 1. This invention creates a hybrid digital twin model that directly links environmental factors, driving behavior, and the cumulative loss of key subsystems. The accumulated loss state variables are then fed back into the underlying performance model. This allows for dynamic and realistic degradation of vehicle performance over time and environmental changes. This allows drivers to experience firsthand the effects of their own operation and the environment, such as vehicle acceleration loss, longer braking distances, or increased jolting, greatly enhancing immersion and battlefield realism.

[0043] 2. By converting every improper driving behavior or high-speed maneuver in adverse conditions into a quantifiable loss of vehicle health, the concepts of "consequences" and "costs" are introduced into simulation training. This expands the driver's focus from the single "search-and-fire" tactical cycle to continuous awareness of vehicle status and equipment health management. This forces the driver to balance mission objectives with equipment loss, deeply aligning driving operations with tactical intent and cultivating the driver's ability to perceive vehicle status and predict potential failures in actual combat.

[0044] 3. In addition, the method of the present invention not only trains drivers but also provides a new training dimension for instructors. Instructors can make more complex tactical decisions based on the real-time wear and tear status feedback from each vehicle in the fleet. By dynamically generating tactical tasks that match the vehicle conditions or setting "dilemma" choices that conflict with time, resources, and space, the method can specifically examine the driver's emergency response skills and tactical decision-making ability, thus realizing a high-level comprehensive tactical training model. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] To more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for describing the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. Those skilled in the art can also derive other drawings based on these drawings without inventive effort. Among them:

[0046] Figure 1 This is an overall flow chart of a tactical simulation training method for wheeled assault vehicles in different environments according to an embodiment of the present invention. DETAILED DESCRIPTION

[0047] To make the above-mentioned objects, features, and advantages of the present invention more clearly understood, the following detailed description of the specific embodiments of the present invention is given in conjunction with the accompanying drawings. It is obvious that the described embodiments are only part of the embodiments of the present invention, not all of them. Based on the embodiments of the present invention, all other embodiments obtained by ordinary persons in this field without creative work should fall within the scope of protection of the present invention.

[0048] In the following description, many specific details are set forth to facilitate a full understanding of the present invention. However, the present invention may also be implemented in other ways different from those described herein. Those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0049] Secondly, the term "one embodiment" or "embodiment" herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in various places throughout this specification does not necessarily refer to the same embodiment, nor does it refer to a separate or selective embodiment that is mutually exclusive of other embodiments.

[0050] The present invention is described in detail with reference to schematic diagrams. For ease of illustration, cross-sectional views of device structures may be partially enlarged and not to scale when describing embodiments of the present invention. Furthermore, the schematic diagrams are merely illustrative and should not limit the scope of the present invention. Furthermore, in actual production, the three-dimensional dimensions of length, width, and depth should be included.

[0051] In the description of the present invention, it should be noted that the terms "upper, lower, inner, and outer" and other references to orientations or positional relationships are based on the orientations or positional relationships shown in the accompanying drawings and are intended solely to facilitate and simplify the description of the present invention. They are not intended to indicate or imply that the devices or components referred to must have, be constructed, or operate in a specific orientation, and therefore should not be construed as limitations on the present invention. Furthermore, the terms "first, second, or third" are used for descriptive purposes only and should not be construed as indicating or implying relative importance.

[0052] In this disclosure, unless otherwise specified or limited, the terms "mounted," "connected," and "connected" should be interpreted broadly. For example, they may refer to fixed, removable, or integral connections. They may also refer to mechanical, electrical, or direct connections, indirect connections through an intermediary, or internal communication between two components. Those skilled in the art will understand the specific meanings of these terms in this disclosure.

[0053] Example 1

[0054] Reference Figure 1 , which is the first embodiment of the present invention, provides a tactical simulation training method for wheeled assault vehicles in different environments, including:

[0055] Specifically, the simulation training system loads preset digital map data containing elevation, surface material and weather information to form a specific simulated battlefield training environment for vehicle interaction;

[0056] It should be explained that elevation refers to the data of the altitude of a certain point on the surface. When these data points are densely distributed in an area, they can together form a three-dimensional digital surface model. In the solution of the present invention, the elevation data is mainly used for the physical and mechanical simulation of the wheeled assault vehicle;

[0057] It should be explained that surface material data, also known as land cover data, is data that describes the types of physical cover on the Earth's surface. Land cover data can be used to divide map areas into different categories, thereby simulating the different environments of the current training battlefield. It is obtained by receiving military remote sensing satellite ground systems.

[0058] It should be explained that digital map data of meteorological information usually refers to a four-dimensional (i.e., three-dimensional space + time) meteorological data field, which is used to provide various meteorological parameters at different geographical locations, altitudes, and times, such as temperature, humidity, air pressure (air density), wind speed, wind direction, and precipitation. Its function and acquisition method are the same as those of surface material data.

[0059] Furthermore, the geographical coordinates of the wheeled assault vehicle in the simulation training are obtained in real time;

[0060] Specifically, the simulation training system receives the signals transmitted by the steering wheel, accelerator, brake pedal and other devices at time point t to obtain the current driver's operation instructions (such as steering wheel angle, accelerator opening, braking force), and then uses the previous time step The position coordinates of the assault vehicle are obtained to obtain various forces exerted on the vehicle by the current simulated battlefield training environment, such as the gravitational component force generated by the slope, such as the tire longitudinal force (driving / braking force) and lateral force (steering force) calculated from the surface material and tire slip rate, such as air resistance; considering the above various forces as input, a system of ordinary differential equations is established, which describes the application of Newton's second law to the assault vehicle. The system of ordinary differential equations is solved using the numerical integration method to obtain the linear acceleration and angular acceleration of the vehicle at time point t; integrating the linear acceleration and angular acceleration once can obtain new velocity and angular velocity, and then performing another integration on this basis, that is, integrating the linear acceleration and angular acceleration twice, can obtain the geographic location coordinates and posture of the assault vehicle at time point t;

[0061] Specifically, the ordinary differential equations can be expressed as:

[0062] ,

[0063] in, It represents the resultant force of all external forces acting on the center of mass of the vehicle (including the engine driving force, braking force and various resistances generated by the environment). It represents the resultant torque of all the torques acting on the vehicle; m represents the curb mass of the vehicle, and I represents the moment of inertia of the vehicle as a rigid body rotating around its center of mass; Expressed as the center-of-mass acceleration, Expressed as angular acceleration;

[0064] It should be noted that since a vehicle is a three-dimensional object that can perform three basic rotations in space (corresponding to the x-axis, y-axis, and z-axis), the moment of inertia around the vehicle's longitudinal axis determines the vehicle's ability to resist roll, the moment of inertia around the vehicle's lateral axis determines the vehicle's ability to resist pitch, and the moment of inertia around the vehicle's vertical axis determines the vehicle's ability to resist yaw (i.e., the vehicle's steering characteristics).

[0065] Furthermore, the simulation training system parses the multi-scale environmental parameter vector of the current location from the digital map data based on the geographic coordinates of the assault vehicle;

[0066] It should be noted that because vehicle wear and performance degradation are essentially the result of continuous and dynamic interaction between the vehicle's internal state and the external environment, it is necessary to convert the static environmental information obtained in real time into a real-time, dynamic process that affects the vehicle.

[0067] Specifically, the simulation training system locates the smallest cube containing the space-time point in the index of the four-dimensional meteorological data field based on the acquired geographic coordinates of the assault vehicle and the current simulation time. The cube is composed of 8 adjacent grid points. Then, in order to obtain the meteorological value of the precise point where the geographic coordinates of the assault vehicle are located, a four-dimensional linear interpolation process is performed to calculate the meteorological data field data at the location and time of the geographic coordinates of the assault vehicle, which represents the large-scale, slowly changing environmental characteristics of the vehicle.

[0068] Specifically, the four-dimensional linear interpolation process is:

[0069] At the corner points of 8 adjacent grid points, linear interpolation is performed 4 times along the longitude direction to obtain 4 intermediate points. Then, linear interpolation is performed twice along the latitude direction on these 4 intermediate points to obtain 2 intermediate points. Then, linear interpolation is performed once along the altitude direction on these 2 intermediate points to obtain 1 intermediate point. Finally, linear interpolation is performed one final time along the time axis on the adjacent intermediate points before and after this time point.

[0070] Specifically, the simulation training system extracts the two-dimensional geographic coordinates of the assault vehicle and converts them into world coordinates based on the acquired geographic coordinates of the assault vehicle. At the same time, it loads the surface cover classification raster map composed of land cover data and converts the world coordinates of the assault vehicle into the pixel coordinates of the raster map. Because the surface cover classification is a discrete category (e.g., 1 = water, 2 = forest) rather than a continuous value, interpolation is not used here. Instead, the pixel value of the pixel coordinate in the raster image is directly read and used as the category of the surface material data, representing the local environment characteristics of the vehicle.

[0071] Specifically, the pixel coordinates of the raster image are converted to:

[0072] ,

[0073] ,

[0074] in, and They represent the world coordinates after the two-dimensional geographic coordinates are converted. Expressed as pixel width, Expressed as pixel width;

[0075] Specifically, the road roughness index is calculated based on the vehicle coordinates, the type of surface material data, and the tire slip rate (obtained through the vehicle's own dynamics).

[0076] Specifically, the calculation process of the road roughness index is as follows:

[0077] Based on the surface material data type (such as grass, sand, or concrete), a sampling rectangular area (for example, a 5m x 2m rectangular area comparable in size to the vehicle body) is defined with the projection of the vehicle's center of mass on the ground as the center. Elevation data is then queried to obtain all elevation data points within the rectangular area. Using the least squares method, an optimal plane is fitted from these elevation data points. This plane represents the slope of the vehicle in the current area. The vertical distance (i.e., residual) from each elevation point to the fitted plane is calculated, and the root mean square (RMS) of all residuals is calculated to reflect the degree of undulation of the road surface relative to its average slope. This RMS value is then normalized (ranging from 0 to 1) to obtain the current road surface inequality index, with values ​​closer to 0 indicating flatness and values ​​closer to 1 indicating bumpiness.

[0078] In addition to the road roughness index, it is also necessary to consider the vehicle's adhesion coefficient and grinding coefficient relative to the road surface, as well as its own filter clogging coefficient. The adhesion coefficient is a dynamic quantity expressed as the ratio of the maximum static friction force that can be generated between the tire and the ground to the normal pressure. It is not a fixed value. This value can be obtained by combining surface material data, tire slip rate, vehicle speed, and tire side slip angle test bench data. The tire side slip angle refers to the angle between the actual movement direction of the tire and the direction of the tire when the vehicle turns, that is, the side slip angle; the grinding coefficient is mainly used to quantify the road's ability to micro-cut and wear the tire tread rubber. Its value can be derived from the temperature and humidity in the four-dimensional meteorological data field combined with surface material data. The filter blockage coefficient does not represent the state of the filter itself, but rather represents the mass concentration of suspended particulate matter generated by vehicle motion in the environment and potentially drawn into the engine air intake per unit time. Its value, in addition to surface material data and the current vehicle speed, also includes the tactical formation position (i.e., when following another vehicle, it inhales a large amount of dust kicked up by the vehicle in front, and the filter of the following vehicle clogs much faster than that of the leading vehicle). This tactical formation position is determined by determining whether there are other friendly vehicles within a certain distance and angle in front of the current vehicle.

[0079] It should be noted that using the road roughness index, adhesion coefficient and abrasive coefficient, and the vehicle's filter clogging coefficient as multi-scale environmental parameter vectors enables drivers to not only experience the real driving behavior of vehicle combat during tactical simulation training, but also consider the vehicle's own implicit losses, thereby deeply linking the driver's driving behavior with tactical decision-making. For example, when an assault vehicle team is conducting an infiltration operation on a dry dirt road, their combat objective is to seize an enemy high ground. If the driver fails to consider the vehicle's own losses in a timely manner, and the vehicle's performance drops by 10% due to filter clogging, it may directly lead to failure in seizing the point.

[0080] Furthermore, a hybrid digital twin model is constructed for each subsystem in the wheeled assault vehicle used for simulation training;

[0081] It should be explained that the hybrid digital twin model is composed of a physical model of the actual device and a machine learning model. The model uses the physical model of the actual device as a framework to ensure that the vehicle's driving behavior conforms to the laws of physics at a macro level, and uses the machine learning model as a component to ensure that the physical model conforms to the laws of physics while being as close to the real environment as possible.

[0082] Specifically, each hybrid digital twin model includes: a physical-based basic vehicle performance model, which is used to obtain the basic performance parameters of the vehicle based on the driver's driving behavior vector; and a data-driven residual compensation model trained with a machine learning algorithm, which is used to obtain compensation values ​​for the basic vehicle performance parameters based on the multi-scale environmental parameter vector and the driving behavior vector.

[0083] Specifically, the driving behavior vector is obtained from the current driver's operation instructions;

[0084] Furthermore, the subsystems in a wheeled assault vehicle consist of a tire subsystem, a suspension subsystem, and a powertrain subsystem;

[0085] Specifically, the tire subsystem is mainly used to calculate the longitudinal force (driving / braking force) and lateral force (steering force) generated when the tire contacts the road (adhesion coefficient);

[0086] Specifically, the physical mechanism of the tire subsystem is based on the basic vehicle performance model. This physical mechanism is based on the tire slip and side deviation theory and is implemented using the classic Fiala tire model.

[0087] Specifically, based on the physical mechanism-based vehicle performance model of the tire subsystem, the driver's driving behavior vector is used as input to calculate the longitudinal slip rate and slip angle of each tire in real time. In the case of longitudinal slip (slip angle = 0), the longitudinal force generated by the tire is a nonlinear function of the slip rate. When the slip rate is small, the longitudinal force generated by the tire has an approximately linear relationship with the longitudinal slip. As the slip rate increases, the longitudinal force generated by the tire reaches a peak (maximum driving / braking force) before beginning to decline. In the case of lateral slip (longitudinal slip rate = 0), the lateral force is a nonlinear function of the slip angle and similarly exhibits a characteristic of decreasing after reaching a peak. When both longitudinal slip rate and slip angle exist, the longitudinal and lateral forces generated by the tire are coupled and calculated using the concepts of a "friction ellipse" or "friction circle" to ensure that their resultant force does not exceed the maximum adhesion between the tire and the road surface. Finally, the model outputs the three-dimensional force vector generated by each tire.

[0088] Specifically, the suspension subsystem is mainly used to calculate the support force and damping force generated by the suspension system (springs and shock absorbers) during vehicle body movement, and to determine the dynamic vertical load on each wheel;

[0089] Specifically, for the physical mechanism underlying the vehicle performance model of the suspension subsystem, the physical mechanism is based on vibration theory and is implemented using a quarter-based vehicle model;

[0090] Specifically, the vehicle suspension is simplified into four independent quarter-vehicle models. Each vehicle model consists of two mass blocks (the sprung mass represents the body, and the unsprung mass represents the wheels and suspension arms) and two connecting elements (the spring and shock absorber represent the suspension, and the spring represents the radial stiffness of the tire). Using Newton's second law, a set of second-order ordinary differential equations is established for the sprung mass and unsprung mass to describe their vertical vibration:

[0091] ,

[0092] ,

[0093] in, is the sprung mass, is the unsprung mass, is expressed as the vertical displacement of the sprung mass, is expressed as the vertical displacement of the unsprung mass, Expressed as the suspension shock absorber damping coefficient, Expressed as the suspension spring stiffness, , Expressed as the vertical force generated by the tire, Expressed as the road roughness index, Expressed as tire radial stiffness, Expressed as tire damping coefficient;

[0094] In addition to vertical vibration, the model can also calculate the load transfer caused by longitudinal acceleration / deceleration and lateral acceleration through algebraic equations. This transfer will be superimposed on the static load to form a dynamic vertical load;

[0095] Specifically, the powertrain subsystem is mainly used to calculate the driving / braking torque ultimately transmitted to the wheels by the engine and transmission system under the current driving behavior;

[0096] Specifically, for the physical mechanism-based vehicle performance model of the powertrain subsystem, the physical mechanism is implemented based on the torque transfer relationship of the transmission system and is implemented using a lookup table engine mapping model;

[0097] Specifically, the model preloads one or more engine universal characteristic maps measured from engine test bench data. This map is a three-dimensional data table whose indexes are engine speed and throttle opening, and the values ​​stored in the table are the corresponding engine output torque. Furthermore, at each time step, the model uses bilinear interpolation to query this three-dimensional data table based on the input engine speed and throttle opening to obtain the current vehicle's base engine torque. The queried engine torque is then transformed through the engine mapping model, taking into account the transmission ratio of the current gear of the transmission, the transmission ratio of the final reducer, and the mechanical efficiency of the driveline, to obtain the wheel drive torque:

[0098] ,

[0099] in, is the wheel driving torque, Expressed as engine output torque, determined by the current engine speed and the depth of the accelerator pedal pressed by the driver Joint decision, Further expressed as , Indicates the current gear ratio of the transmission. Indicates the main reducer transmission ratio, Expressed as the total mechanical efficiency of the transmission system, Expressed as wheel speed;

[0100] It should be noted that by considering the core indicators of each subsystem in the above-mentioned wheeled assault vehicle, the basic vehicle performance model can clearly and accurately describe the current performance status of the vehicle;

[0101] Furthermore, based on the above basic vehicle performance model, a data-driven residual compensation model corresponding to the model is constructed for each subsystem in the wheeled assault vehicle;

[0102] Specifically, the residual compensation model is trained by a machine learning algorithm that uses a gradient boosting decision tree. The core idea of ​​the gradient boosting decision tree is to iteratively construct a series of weak learners (i.e., decision trees). Each new decision tree is dedicated to fitting and correcting the accumulated prediction residuals of all previous decision trees.

[0103] Specifically, the multi-scale environmental parameter vector and the driving behavior vector are spliced ​​together, and the splicing result is used as the input of the residual compensation model. The splicing result tells the residual compensation model "under what kind of environment and with what kind of driving behavior, how much deviation will the vehicle performance model produce"; each split node in each weak learner (i.e., decision tree) represents a judgment on the multi-scale environmental parameter vector and the driving behavior vector, that is, a simplified ternary question "What is the operation of this parameter? What is the value? What is the current environment?"; when the first decision tree is output, the total residual of the first decision tree is obtained by traversing the residuals of all nodes under the tree, where The residual refers to the difference between the predicted value given by the node and the actual input value. Based on this difference, the residuals of all nodes under the second tree are traversed in the same way. And so on. The goal of each decision tree is to fit and correct based on the residual of the previous decision tree until the newly added decision tree can no longer significantly reduce the error of the model (that is, the model performance can no longer be improved on the validation set) and the maximum number of iterations of the model is reached. Then, the total residual value of each decision tree is weighted and summed to obtain the residual compensation amount. The weight is the learning rate of the model, which is initialized with a small value and can be adjusted according to the complexity of the input value, such as the learning rate of 1e-4.

[0104] In addition, the training method of gradient boosting decision tree is completed offline;

[0105] Specifically, a data set is formed using digital map data of original elevation, surface material, and meteorological information. Data on this data set is randomly collected to form synchronized data points. Each set of synchronized data points contains real environment data, real driving data, and real subsystem performance data. Based on the processing of these data in the aforementioned basic vehicle performance model, each set of synchronized data points is used as input. The residual of each set of synchronized data points is used as the feature of the synchronized data point through the corresponding residual compensation model. The real environment data and real driving data in the synchronized data points are spliced ​​as the features of the synchronized data points to construct a structured training data set consisting of {feature, label} pairs containing several samples. The labels of the generated synchronized data points are then randomly divided into a training set, a validation set, and a test set in a ratio of 8:1:1. The residual compensation model is trained using the training set. The learning rate of the offline training is set to search in the range of [0.01, 0.1], and the maximum depth of the decision tree is searched in the range of [3, 10]. The residual compensation model completed by offline training is converted into a standard model format (such as .pkl) for storage.

[0106] It should be noted that vehicle performance models based on physical mechanisms are bound to produce systematic deviations when faced with complex, coupled real-world environments. Although these can be addressed by establishing complex physical equations, the model's computational cost is too high, making it unsuitable for real-time simulation training systems. However, establishing a residual compensation model corresponding to the basic vehicle performance model not only replaces the complex physical equations but also compensates for the core indicators in the physical model by combining driving behavior and environmental factors.

[0107] Furthermore, within each time step, the vehicle's accumulated loss state variables are updated in real time based on the residual-compensated subsystem core parameters.

[0108] Specifically, at the beginning of the simulation training task, one or more cumulative loss state variables are initialized for each key component whose vehicle loss needs to be tracked. Their initial values ​​are all set to 0, indicating that the vehicle is in a brand-new, intact state. Among the key components, the three subsystems of the wheeled assault vehicle are mainly considered. In addition, the longitudinal slip rate of the i-th tire is obtained from the hybrid digital twin model representing the tire subsystem, and the dynamic vertical load applied to the i-th tire is obtained from the hybrid digital twin model representing the suspension subsystem. The current road surface grinding coefficient is obtained from the multi-scale environmental parameter vector. The wear energy dissipation rate per unit time is calculated based on the current vehicle speed:

[0109] ,

[0110] in, represents the wear energy dissipation rate of the i-th tire; Expressed as the dynamic vertical load of the i-th tire. Since friction is proportional to the normal pressure (i.e., vertical load) according to Coulomb's law, by considering the dynamically changing vertical load, the wear calculation of the inner and outer tires is more realistic when the vehicle is cornering or bumping; Expressed as the longitudinal slip rate of the i-th tire, is represented by the sideslip angle of the i-th tire, Expressed as the instantaneous speed of the vehicle;

[0111] It should be noted that by calculating the wear energy dissipation rate per unit time, it is possible to quantify the rate at which energy is dissipated in the form of frictional heat due to the relative sliding between the tire and the road surface at the current time t. This rate can be regarded as the direct physical cause of tire wear rather than the direct amount of wear.

[0112] Furthermore, the wear energy dissipation rate is converted into actual wear depth increments to simulate the wear of the tire throughout its life cycle:

[0113] ,

[0114] in, Expressed as the wear depth increment of the i-th tire, Expressed as the tire material wear coefficient, an empirical constant determined by the test bench, Expressed as the road surface grinding coefficient, Expressed as time steps;

[0115] It should be noted that the test bench determination method for the wear coefficient of tire materials is obtained through offline calibration in the following steps: Step 1, prepare tire tread rubber samples of standard size and samples of various typical road surfaces (such as cement, asphalt, and sand); Step 2, use a pin-on-disc friction and wear tester to perform a wear test of the rubber sample on the road surface sample for a specified mileage under a set normal pressure (simulating vehicle load) and relative sliding speed; Step 3, measure the mass loss of the sample and combine it with the The calculation formula is reversed Step 4: Repeat steps 1 to 4 and test under various conditions to obtain the tire material wear coefficient.

[0116] Furthermore, the difference between the vertical displacement of the unsprung mass and the vertical displacement of the sprung mass is obtained from the hybrid digital twin model representing the suspension subsystem, that is, The difference is multiplied by an effective stress-to-stroke conversion coefficient to represent the stress value. The effective stress-to-stroke conversion coefficient is obtained through offline calibration using the following steps: First, a finite element analysis (FEA) model of a key suspension component (such as a spring or lower control arm) is established. A series of static displacement loads from 0 to maximum stroke are then applied to the FEA model. The maximum principal stress value at each key stress monitoring point (for example, a high-stress area identified by a stress contour map) corresponding to each displacement is recorded. Finally, a linear regression analysis is performed on this stress monitoring point. The slope of the line generated by this regression analysis is the effective stress-to-stroke conversion coefficient. Based on this stress value, the rain flow counting method is used to identify each equivalent stress cycle. Combined with the SN curve (the SN curve is an inherent material property measured by standard fatigue testing and describes the number of cycles a material can withstand without fatigue damage under different stress amplitudes) of key load-bearing suspension components (for example, the root weld of the lower control arm, which is most susceptible to fatigue fracture), the damage degree of the vehicle suspension is estimated:

[0117] ,

[0118] in, is the fatigue damage degree caused by the jth stress cycle, which represents the percentage of the total fatigue life of the key suspension components consumed by the jth stress cycle identified by the rainflow counting method; is the number of allowable cycles under specific stress conditions, is the stress amplitude obtained from the SN curve, which represents the half height of the jth equivalent stress cycle decomposed from the real-time stress history by the rainflow counting method; is the mean stress, and Similarly, it represents the midline position of the jth equivalent stress cycle;

[0119] Furthermore, the engine output torque (i.e. the current engine speed) is obtained from the hybrid digital twin model representing the powertrain subsystem. and the depth of the accelerator pedal pressed by the driver ), and obtain the filter blockage coefficient from the multi-scale environmental parameter vector. By considering the filtration efficiency of the filter, the particulate matter adsorption mass is calculated within the time step to obtain the potential damage to the vehicle power:

[0120] ,

[0121] in, represents the blockage mass increment of the kth filter in the time step; Indicates the instantaneous intake flow of the engine, which is obtained from the engine output torque; It is expressed as the ambient air particulate matter concentration, which is obtained by the multi-scale environmental parameter vector based on the current environmental assessment; Indicates the filtration efficiency of the filter, which is affected by the specific filter model;

[0122] It should be noted that by considering the vehicle's cumulative wear and tear (including but not limited to the aforementioned wear and tear values), every improper driving behavior or every high-speed maneuver in adverse conditions is converted into a real vehicle health value for deduction. This introduces the concepts of "consequences" and "costs" of battlefield decision-making, forcing drivers to manage the vehicle's "life" as they would their own, achieving true single-vehicle tactical simulation training. At the same time, instructors can decide whether to continue the assault, dispatch some vehicles for tactical maintenance, or adjust the entire formation's route and speed based on the real-time wear and tear status reported by the drivers of each vehicle in the fleet. This also elevates the dimension of simulation training from the driver's single vehicle operation to the level of comprehensive fleet resource management.

[0123] Furthermore, by feeding the accumulated vehicle loss values ​​obtained above back into the corresponding hybrid digital twin model to modify the parameters in the base vehicle performance model, the driver can personally experience the feedback effect of their driving behavior on vehicle performance. For example, when tires are severely worn, their grip will be significantly reduced, and the driver will feel that the vehicle is more prone to slipping and has a longer braking distance. When the suspension is fatigued, its shock absorption effect will deteriorate, and the driver will feel more severe vehicle bumps and reduced stability at high speeds. When the filter is severely clogged, the driver will feel that the vehicle has a weak acceleration.

[0124] Furthermore, considering the status monitoring and graded warning of vehicle subsystems, the accumulated loss value of the vehicle can be preset as a three-level threshold table:

[0125] Level 1: Yellow warning threshold (for example, if the left front tire is worn more than 60%, it indicates that the left front tire is currently severely worn);

[0126] Level 2: Red alert threshold (for example, if the filter blockage coefficient exceeds 0.8, it indicates that the power system efficiency has seriously decreased and there is a risk of stalling);

[0127] Level 3: Critical failure threshold (for example, if the suspension fatigue reaches 1.0, it indicates that the current vehicle suspension system is damaged and the vehicle is out of control);

[0128] Furthermore, tactical tasks are formulated based on the current vehicle's comprehensive cumulative losses, and the current driver's driving skills are examined. For example, if the current vehicle's maneuverability is severely degraded (i.e., the tires or suspension are nearing their limits), but the power system is intact and still has strong acceleration capabilities, then based on the vehicle's condition, the task can be to conduct a short reconnaissance at a high-value observation point that can only be reached through rugged terrain. The success of the task directly depends on the vehicle's weakest link (the suspension), which forces the driver to make risk trade-offs in the route selection for the task execution. For example, the instructor issues an urgent task, "Because your crew has penetrated deep behind enemy lines, you are now tasked to advance to the "Eagle's Nest" observation point at coordinates (XXX, YYY) within 15 minutes to confirm enemy movements. Special reminder: To reach the "Eagle's Nest," you must traverse a 10-kilometer stretch of rugged terrain. Please pay attention to the current vehicle's suspension system."

[0129] It should be noted that this task is not an "impossible task". The task settings are adjusted based on the current vehicle damage value to ensure the feasibility of the task settings. For example, considering various routes, if the suspension fatigue is 0.65, then the length of the rough terrain to the observation point is 8 to 12 kilometers based on distance measurement, which just makes the fatigue close to 0.95 (critical value 1); if the suspension fatigue is 0.75, then the length of the rough terrain will be shortened to 7 to 10 kilometers based on distance measurement, ensuring that each triggered task is "tailor-made" for the current driver;

[0130] Specifically, the distance measurement formula is as follows:

[0131] ,

[0132] in, It is the length of the terrain measured by the system. The wear threshold is a preset, dimensionless constant that represents the degree of cumulative wear of key components (such as the suspension system) expected to be achieved after the vehicle completes the journey. This value is usually set very close to the critical failure point of the component (for example, 0.95, while the critical value is 1.0) to test the driver's driving skills. is the accumulated loss value of the current key components; Expressed as the unit distance loss coefficient, that is, the incremental loss of key components per kilometer;

[0133] It should be noted that the unit distance loss coefficient is obtained by inputting the current driver's driving behavior and the driver's planned path, dividing the path into several 1-kilometer standard sections. For each standard section, based on the multi-scale environmental parameter vector combined with the hybrid digital twin model, the cumulative loss increment of key components on each 1-kilometer path section under different environments is obtained, which is the unit distance loss coefficient.

[0134] Specifically, the current driver's route, every driving action (such as throttle, brake, steering, speed control, etc.) and the wear and tear values ​​of the vehicle's key components are recorded and used as a basis for review by the simulation training system;

[0135] In addition to testing the driver's driving skills, the game can also be used to deliberately create dilemmas or even trilemmas to test the driver's tactical decision-making. There are three types of game playing:

[0136] Time Conflict: If the simulation training system calculates that a tire change requires 8 minutes, while a high-value reconnaissance mission has only 10 minutes left, the system will display a warning message: "Warning: The right rear tire is severely worn, the distance limit is estimated to last 15 minutes, and the reconnaissance mission must reach the target point within 10 minutes." This forces the driver to make a decision: perform maintenance immediately (which will result in missing the mission) or continue on (which will result in a high risk of a tire blowout).

[0137] Resource conflict type: There is only one spare tire on the vehicle, but the wear of both left and right tires is close to the limit. The system will prompt: Warning, both front wheels are severely worn, but there is only one spare tire on the vehicle. Please choose to replace the left or right wheel.

[0138] Spatial Conflict: The simulation training system predicts that the vehicle must be maintained within 20 minutes, but based on the map and enemy analysis, it will be on an open plain with no cover for the next 25 minutes. This forces the crew to make a decision between "risking exposure on the plain for maintenance" and "forcing their way to a safe zone but potentially breaking down midway."

[0139] Furthermore, the current equipment preservation score is calculated based on the driver's driving skills and tactical decisions to assess the driver's maintenance management capabilities;

[0140] Specifically, the current equipment preservation score is expressed as:

[0141] ,

[0142] in, It is represented as the equipment preservation score, ranging from [0 to 100]. The higher the score, the better the driver's driving behavior and tactical decision-making control of equipment loss. N is the sum of the key components of the vehicle. For example, if 3 tires, 2 suspensions and 1 filter are monitored, then N = 6. It is expressed as the degree of impact of the damage of the jth component on the comprehensive combat effectiveness of the vehicle. It is a weight value. This weight value is obtained by military experts and vehicle maintenance experts based on the failure mode, impact and criticality analysis results of this type of wheeled assault vehicle and the requirements of typical combat scenarios. Expressed as the real-time cumulative loss value of the j-th component; represents the critical failure threshold of the jth component;

[0143] In addition, to ensure that the equipment preservation score serves tactical training guidance and allows instructors to conduct accurate and well-founded replay analysis, the simulation training system automatically generates a multi-window interactive replay after the simulation training is completed. This interactive replay interface includes a tactical map panel: used to display the complete trajectory of the vehicle, with the trajectory line colored according to the equipment preservation score value at the time and highlighted with a special icon on the map; a data chart panel: with time as the horizontal axis, synchronously displaying the equipment preservation score curve and the curve of each key component loss; a playback view panel: used to replay training videos of any time period from the driver's perspective or a third-party perspective;

[0144] Specifically, through the above steps, instructors can use this powerful review tool to provide the following precise and effective teaching guidance: For example, the instructor can guide the driver: "Look, at 12 minutes and 35 seconds, you chose to cross this gravel slope to take a shortcut. Although you saved 30 seconds, your equipment maintenance score decreased, and the suspension system accumulated 35% of the total fatigue damage. It was this damage that caused the vehicle to tilt too much when you were avoiding enemy fire from behind, and eventually overturned.

[0145] Those skilled in the art will appreciate that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present application may take the form of a fully hardware embodiment, a fully software embodiment, or an embodiment combining software and hardware. Furthermore, the present application may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code. The solutions in the embodiments of the present application may be implemented using various computer languages, such as the object-oriented programming language Java and the interpreted scripting language JavaScript.

[0146] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 a process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0147] These computer program instructions may also be stored in a computer readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 a process or multiple processes and / or boxes Figure 1 The function specified in one or more boxes.

[0148] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operational steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing the instructions executed on the computer or other programmable device for implementing the process. Figure 1 a process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0149] Although the preferred embodiments of the present application have been described, those skilled in the art may make additional changes and modifications to these embodiments once they have learned the basic creative concept. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications that fall within the scope of the present application.

[0150] Obviously, those skilled in the art may make various changes and modifications to this application without departing from the spirit and scope of this application. Thus, if these modifications and variations of this application fall within the scope of the claims of this application and their equivalents, this application is intended to include these modifications and variations.

Claims

1. A tactical simulation training method for wheeled assault vehicles in different environments, characterized in that: include: Obtain the multi-scale environmental parameter vector of the current position of the wheeled assault vehicle in simulation training; Calculating compensated performance parameters of at least one key subsystem of the wheeled assault vehicle using a hybrid digital twin model, wherein the hybrid digital twin model is composed of a physical mechanism-based basic performance model and a data-driven residual compensation model; The key subsystems include at least a tire subsystem, a suspension subsystem and a powertrain subsystem; The tire subsystem is mainly used to calculate the longitudinal force and lateral force generated when the tire contacts the road surface; The suspension subsystem is mainly used to calculate the support force and damping force generated by the suspension system during vehicle body movement, and to determine the dynamic vertical load on each wheel; The powertrain subsystem is mainly used to calculate the driving / braking torque ultimately transmitted to the wheels by the engine and transmission system under the current driving behavior; The basic performance model calculates the basic performance parameters of the key subsystem according to the driving behavior vector; The residual compensation model calculates a residual compensation value for the basic performance parameter based on the multi-scale environmental parameter vector and the driving behavior vector; Calculating the accumulated loss state variable of the key subsystem based on the compensated performance parameter of the key subsystem; The step of calculating the accumulated loss state variable of the key subsystem includes: For the tire subsystem, calculating the increment of tire wear depth according to the dynamic vertical load, longitudinal slip rate, sideslip angle and road abrasion coefficient of the tire; For the suspension subsystem, the fatigue damage degree of key suspension components is calculated based on the dynamic travel of the suspension using the rain flow counting method and the preset SN curve; For the powertrain subsystem, calculating the filter blockage mass increment based on the instantaneous intake air flow of the engine and the ambient air particulate matter concentration; The performance status of the key subsystems in the simulation training is dynamically updated based on the feedback of the accumulated loss state variable to the basic performance model.

2. The wheeled assault vehicle tactical simulation training method for different environments as claimed in claim 1, characterized in that: The steps of obtaining the multi-scale environmental parameter vector include: Load preset digital map data containing elevation, surface material and weather information; Acquiring the geographical location coordinates of the wheeled assault vehicle in real time; According to the geographic location coordinates, parameters including at least a road surface roughness index, a road surface adhesion coefficient, a road surface abrasion coefficient, and a filter clogging coefficient are parsed from the digital map data to form the multi-scale environmental parameter vector.

3. The wheeled assault vehicle tactical simulation training method for different environments as claimed in claim 1, characterized in that: The residual compensation model is a machine learning model obtained by offline training of the gradient boosting decision tree algorithm; The input of the model is a feature vector formed by splicing the multi-scale environmental parameter vector and the driving behavior vector, and the output is the residual compensation value.

4. The wheeled assault vehicle tactical simulation training method for different environments as claimed in claim 3 is characterized in that: The offline training method of the gradient boosting decision tree algorithm includes: Collect synchronized data points containing real-world environment data, real-world driving data, and real-world subsystem performance data; For each synchronous data point, inputting the actual driving data into the basic performance model to obtain a theoretical prediction value; Calculating the difference between the actual subsystem performance data and the theoretical prediction value as a training label; splicing the real environment data and the real driving data as training features; The gradient boosting decision tree model is trained using the generated training dataset consisting of {feature, label} pairs.

5. The wheeled assault vehicle tactical simulation training method for different environments as claimed in claim 1 is characterized in that: Also includes: Presetting at least one early warning threshold value of the cumulative loss state variable; When any accumulated loss state variable reaches its corresponding warning threshold, graded warning information is provided to the driver.

6. The wheeled assault vehicle tactical simulation training method for different environments as claimed in claim 1, characterized in that: Also includes: According to the current cumulative loss state of the wheeled assault vehicle, a tactical task with constraints that matches the current vehicle performance state is dynamically generated.

7. The wheeled assault vehicle tactical simulation training method for different environments as claimed in claim 1 is characterized in that: Also includes: Based on the current cumulative loss status of the wheeled assault vehicle, mission time limit and available maintenance resources, at least one time-conflict, resource-conflict or space-conflict tactical decision-making game scenario is constructed to force the driver to make a trade-off between equipment maintenance and mission execution.

8. The wheeled assault vehicle tactical simulation training method for different environments as claimed in claim 1 is characterized in that: Also includes: Calculate and update a comprehensive equipment preservation score in real time based on the real-time cumulative loss value of each key subsystem and its preset impact weight on the combat effectiveness of the entire vehicle; After the simulation training is completed, an interactive review interface is generated that includes the vehicle's driving trajectory, equipment preservation score curve, and loss curves of key vehicle components for training evaluation and teaching analysis.

Images