Attitude control method and system for mixed target

Abstract

The invention provides an attitude control method and system for a mixed target. The method comprises the following steps: predicting a scene change trend by using a pre-trained neural network model according to topographic data and temperature and humidity data to obtain a scene change trend vector, and adjusting parameters of a target hybrid model according to the scene change trend vector; calculating an air pressure target value through the target hybrid model after parameter adjustment; according to the air pressure target value, the gas injection or release rate of the two-way pressure adjusting equipment of the mixed target is controlled until the fluctuation amplitude of the air pressure actual value after pressure adjustment is smaller than a preset fluctuation amplitude threshold value and the stable duration is larger than a preset duration threshold value, and the two-way pressure adjusting equipment comprises an air pump and a pressure release valve; and controlling the attitude of the mixed target according to the actual air pressure value after the control is completed. According to the technical scheme provided by the invention, the attitude of the mixed target is controlled through the actual air pressure value of the mixed target, and precision and stability can be realized in a complex environment.

Description

Technical Field

[0001] This application relates to the technical field of target attitude control, and particularly to an attitude control method and system for a hybrid target. Background Art

[0002] In military training, weapon testing, or dynamic simulation scenarios, the coordinated control of a movable entity target vehicle and a hybrid target needs to meet the adaptive adjustment requirements in complex environments, especially for inflatable targets. Specifically, the target needs to adapt to the realism requirements of different training scenarios; at the same time, it is required to ensure the stability and durability after air pressure adjustment.

[0003] A typical existing solution for such requirements is an environment prediction control model based on a single neural network. This solution collects environmental data such as terrain, temperature, and humidity, uses a pre-trained recurrent neural network or convolutional neural network to predict the dynamic change trend of the environment, and directly maps it to attitude data.

[0004] The main defect of the existing solution is the lack of generalization ability in a dynamic environment. When a single neural network model processes multi-modal coupled data, it is difficult to simultaneously take into account the rule reasoning ability of fuzzy logic and the global optimization characteristics of non-linear optimization, resulting in the prediction results being too sensitive to control parameters. In addition, the existing solution is prone to overshoot or response lag when facing complex environmental disturbances, and it is difficult to meet the requirements of precise attitude control. Summary of the Invention

[0005] This application provides an attitude control method and system for a hybrid target to solve the problem in the prior art that in a dynamic environment, overshoot or response lag is likely to occur when facing complex environmental disturbances, resulting in poor attitude control accuracy.

[0006] In a first aspect, this application provides an attitude control method for a hybrid target, including: Obtaining the moving speed data of a movable entity target vehicle, the hardness level parameter data and the actual air pressure value of the hybrid target, and the terrain data and temperature and humidity data of the environment where the movable entity target vehicle is located; According to the terrain data and temperature and humidity data, predicting the scene change trend through a pre-trained neural network model to obtain a scene change trend vector, and adjusting the parameters of a target hybrid model according to the scene change trend vector, where the target hybrid model is a hybrid model integrating a fuzzy logic module and a non-linear optimization module; Calculating the target air pressure value through the target hybrid model with adjusted parameters based on the moving speed data, the hardness level parameter data, the terrain data, and the temperature and humidity data; According to the target air pressure value, control the gas injection or release rate of the two-way pressure regulating device of the hybrid target until the fluctuation range of the actual air pressure value after pressure regulation is less than the preset fluctuation range threshold and the stable duration is greater than the preset duration threshold. After the control is completed, control the attitude of the hybrid target. The two-way pressure regulating device includes an air pump and a pressure relief valve.

[0007] Optionally, calculating the target air pressure value through the target hybrid model after parameter adjustment based on the moving speed data, the hardness level parameter data, the terrain data, and the temperature and humidity data includes: Align the sampling timestamps of the moving speed data and the hardness level parameter data to generate a multi-feature matrix; Perform second-order polynomial basis expansion on the multi-feature matrix to generate an expanded feature matrix, perform local response normalization on the environmental feature tensor to generate a normalized tensor, and calculate a differential feature vector based on the time series of the historical air pressure values. The environmental feature tensor is generated according to the terrain data and the temperature and humidity data; Perform weighted splicing on the expanded feature matrix, the normalized tensor, and the differential feature vector to form a hybrid feature vector; Perform non-linear mapping on the hybrid feature vector through the target hybrid model after parameter adjustment to obtain the target air pressure value.

[0008] Optionally, performing second-order polynomial basis expansion on the multi-feature matrix to generate an expanded feature matrix, performing local response normalization on the environmental feature tensor to generate a normalized tensor, and calculating a differential feature vector based on the time series of the historical air pressure values includes: For each feature dimension in the multi-feature matrix, calculate the original feature value on the feature dimension, the product term of two feature dimensions, and the square term of a single feature dimension, and splice the original feature value, the product term, and the square term by column to generate an expanded feature matrix; For each channel of the environmental feature tensor, select the sum of the squares of the activation values of adjacent m channels as the normalization base number, weight the normalization base number with a preset channel attenuation factor, divide the activation value of the current channel by the weighted normalization base number to obtain a normalized value, and perform non-linear compression on the normalized value through the hyperbolic tangent function to generate a normalized tensor; With a fixed window length, intercept multiple sequences of historical air pressure values within multiple windows from the time series of historical air pressure values, calculate the absolute difference sequence between adjacent time points of the historical air pressure values within each window, perform moving average filtering on the absolute difference sequence to obtain the filtered sequence corresponding to each window; perform cumulative summation on the filtered sequences corresponding to all windows to generate a differential feature vector.

[0009] Optionally, the weighted concatenation of the extended feature matrix, the normalized tensor, and the differential feature vector to form a hybrid feature vector includes: Assigning a weight coefficient related to the feature dimension, a weight coefficient related to the number of channels, and a weight coefficient with a fixed value to the extended feature matrix, the normalized tensor, and the differential feature vector respectively; Performing Hadamard product operations on the extended feature matrix, the normalized tensor, and the differential feature vector respectively according to the weight coefficient related to the feature dimension, the weight coefficient related to the number of channels, and the weight coefficient with a fixed value to obtain a weighted extended feature matrix, a weighted normalized tensor, and a weighted differential feature vector; Performing format conversion on the weighted extended feature matrix and the weighted normalized tensor respectively to obtain an extended feature vector and a normalized vector; Concatenating the elements at the same positions in the extended feature vector, the normalized vector, and the weighted differential feature vector according to a preset concatenation order to form a hybrid feature vector.

[0010] Optionally, the non - linear mapping of the hybrid feature vector by the target hybrid model with adjusted parameters to obtain the air pressure target value includes: Extracting a first sub - vector related to terrain and target hardness and a second sub - vector related to environment and moving speed from the hybrid feature vector; Inputting the first sub - vector and the second sub - vector into a fuzzy logic module and a non - linear optimization module respectively. The fuzzy logic module calculates a fuzzy inference result according to a predefined fuzzy logic rule base and in combination with the weight parameters of the adjusted membership function, and converts the fuzzy inference result into a first air pressure compensation value. The non - linear optimization module calculates an air pressure reference value based on the penalty factor and iteration step size of the adjusted non - linear optimization algorithm; Superimposing the first air pressure compensation value and the air pressure reference value to obtain the air pressure target value.

[0011] Optionally, according to the terrain data and temperature - humidity data, predicting the scene change trend through a pre - trained neural network model to obtain a scene change trend vector, and adjusting the parameters of the target hybrid model according to the scene change trend vector, includes: Discretizing the terrain data into a terrain elevation matrix according to a preset grid unit, mapping the temperature - humidity data into a temperature - humidity distribution tensor according to spatial coordinates, and performing cross - modal concatenation on the terrain elevation matrix and the temperature - humidity distribution tensor to generate an environmental feature tensor; Input the environmental feature tensor into a pre-trained neural network model, and combine the historical environmental feature tensors within the time sliding window to predict the scene change trend, and output a scene change trend vector representing the topographic dynamic evolution and climate fluctuation intensity; According to the terrain-related components in the scene change trend vector, adjust the weight parameters of the membership functions in the fuzzy logic rule base corresponding to the fuzzy logic module, and according to the climate-related components in the scene change trend vector, adjust the penalty factor and iteration step size of the non-linear optimization algorithm adopted by the non-linear optimization module.

[0012] Optionally, the adjusting the weight parameters of the membership functions in the fuzzy logic rule base corresponding to the fuzzy logic module according to the terrain-related components in the scene change trend vector, and adjusting the penalty factor and iteration step size of the non-linear optimization algorithm adopted by the non-linear optimization module according to the climate-related components in the scene change trend vector includes: Analyze the topographic dynamic components in the scene change trend vector to extract the slope change rate and elevation fluctuation intensity, and analyze the climate fluctuation components in the scene change trend vector to extract the temperature and humidity mutation direction and wind speed correlation intensity; According to the slope change rate and the elevation fluctuation intensity, respectively adjust the weight parameters of the membership function of the first rule and the weight parameters of the membership function of the second rule in the fuzzy logic rule base. The first rule is the rule between slope and hardness compensation, and the second rule is the rule between terrain flatness and air pressure stability; According to the temperature and humidity mutation direction and the wind speed correlation intensity, respectively adjust the penalty factor and iteration step size of the non-linear optimization algorithm.

[0013] In a second aspect, the present application provides a posture control system for a hybrid target, including: An acquisition module, configured to acquire the moving speed data of the movable entity target vehicle, the hardness level parameter data and the actual air pressure value of the hybrid target, and the terrain data and temperature and humidity data of the environment where the movable entity target vehicle is located; An adjustment module, configured to predict the scene change trend through a pre-trained neural network model according to the terrain data and temperature and humidity data to obtain a scene change trend vector, and adjust the parameters of the target hybrid model according to the scene change trend vector. The target hybrid model is a hybrid model integrating a fuzzy logic module and a non-linear optimization module; A calculation module, configured to calculate the target air pressure value through the target hybrid model with adjusted parameters based on the moving speed data, the hardness level parameter data, the terrain data and the temperature and humidity data; An adjustment module is configured to control the gas injection or release rate of the two-way pressure adjustment device of the hybrid target according to the target air pressure value until the fluctuation range of the actual air pressure value after pressure adjustment is less than a preset fluctuation range threshold and the stable duration is greater than a preset duration threshold. After the control is completed, the attitude of the hybrid target is controlled according to the actual air pressure value. The two-way pressure adjustment device includes an air pump and a pressure relief valve.

[0014] In a third aspect, the present application provides a computing device, including a processing component and a storage component; the storage component stores one or more computer instructions; the one or more computer instructions are used to be called and executed by the processing component to implement a method for controlling the attitude of a hybrid target as described in any one of the first aspects.

[0015] In a fourth aspect, the present application provides a computer storage medium storing a computer program, which when executed by a computer, implements a method for controlling the attitude of a hybrid target as described in any one of the first aspects.

[0016] In the present application, a method for controlling the attitude of a hybrid target is provided. The method includes: obtaining the moving speed data of a movable entity target vehicle, the hardness level parameter data of the hybrid target, and the actual air pressure value, as well as the terrain data and temperature and humidity data of the environment where the movable entity target vehicle is located; predicting the scene change trend through a pre-trained neural network model according to the terrain data and temperature and humidity data to obtain a scene change trend vector, and adjusting the parameters of a target hybrid model according to the scene change trend vector. The target hybrid model is a hybrid model integrating a fuzzy logic module and a non-linear optimization module; calculating a target air pressure value through the target hybrid model with adjusted parameters based on the moving speed data, the hardness level parameter data, the terrain data, and the temperature and humidity data; controlling the gas injection or release rate of the two-way pressure adjustment device of the hybrid target according to the target air pressure value until the fluctuation range of the actual air pressure value after pressure adjustment is less than a preset fluctuation range threshold and the stable duration is greater than a preset duration threshold. The two-way pressure adjustment device includes an air pump and a pressure relief valve.

[0017] Through the dynamic environment perception and hybrid intelligent regulation mechanism, this application realizes the precise control of the hardness of the hybrid target in complex scenarios, and further realizes the precise control and stable maintenance of the attitude of the hybrid target. Based on the neural network to predict the trend of scene changes and dynamically adjust the parameters of the fuzzy logic-nonlinear optimization hybrid model, the robustness of the system to terrain mutations and temperature / humidity fluctuations is enhanced, and the sensitivity of control parameters is reduced; multi-source data such as moving speed, target hardness, terrain, and temperature / humidity are fused, and the target air pressure value is calculated through the target hybrid model, taking into account fuzzy rule reasoning and nonlinear global optimization during the calculation process; based on the dynamic gas injection / release rate regulation of the two-way pressure regulating device, combined with the preset fluctuation threshold and stable duration constraint, ensure that the actual air pressure value converges quickly and remains stable for a long time, avoiding the overshoot or lag problems of the traditional proportional-integral-derivative controller.

[0018] Furthermore, through multi-source time-series data fusion and feature engineering optimization, the precise prediction of the target air pressure value is realized. The sampling timestamps of the moving speed and hardness level data are aligned to generate a multi-feature matrix, and the nonlinear representation ability is enhanced through second-order polynomial basis expansion; the noise interference of the terrain / temperature / humidity tensor is suppressed based on local response normalization; a windowed absolute difference sequence is extracted from the historical air pressure values, and a differential feature vector is generated through moving average filtering and accumulation; the extended feature matrix, normalized tensor, and differential feature vector are weighted and spliced into a hybrid feature vector, and the target air pressure value is generated by inputting it into the fuzzy-nonlinear optimization hybrid model with adaptive parameters. Through multi-dimensional feature fusion and dynamic feature optimization, the air pressure control accuracy and stability in complex scenarios are improved. The second-order polynomial expansion enhances the modeling ability of the model for the nonlinear coupling relationship between the moving speed and hardness parameters; local response normalization and hyperbolic tangent compression reduce the interference of local mutations in the environmental feature tensor and improve the generalization ability; windowed differential feature extraction accurately captures the dynamic change trend of the historical air pressure values, combined with moving average filtering to suppress high-frequency noise; the hybrid feature vector fuses multi-scale information, enabling the fuzzy logic and nonlinear optimization modules to cooperate and play their advantages, and finally realizing the rapid convergence and long-term stable control of the target air pressure value.

[0019] These aspects or other aspects of this application will be more clearly understood in the following description of the embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] In order to more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the following will briefly introduce the drawings required for the description of the embodiments or the prior art. Obviously, the drawings in the following description are some embodiments of this application. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0021] Figure 1 It is a flowchart of an attitude control method for a hybrid target provided by an embodiment of this application. Figure 2 A structural schematic diagram of an attitude control system for a hybrid target provided by an embodiment of the present application; Figure 3 A structural schematic diagram of a computing device provided by an embodiment of the present application. Specific embodiments

[0022] In order to enable those skilled in the art to better understand the solution of the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present application.

[0023] In some processes described in the specification, claims and above-mentioned drawings of the present application, a plurality of operations appear in a specific order. However, it should be clearly understood that these operations may not be executed in the order in which they appear in this article or may be executed in parallel. The serial numbers of the operations, such as 11, 12, etc., are only used to distinguish different operations, and the serial numbers themselves do not represent any execution order. In addition, these processes may include more or fewer operations, and these operations may be executed in sequence or in parallel. It should be noted that the descriptions such as "first" and "second" in this article are used to distinguish different messages, devices, modules, etc., do not represent a sequence, nor do they limit that "first" and "second" are of different types.

[0024] The technical solutions in the embodiments of the present application will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative efforts belong to the scope of protection of the present application.

[0025] To solve the problems that the prior art is prone to overshoot or response lag when facing complex environmental disturbances and it is difficult to meet the requirements of precise attitude control, the present application provides an attitude control method for a hybrid target. This method adopts the following concept: by collecting multi-source data such as the moving speed of the target vehicle, the hardness of the target, the environmental terrain, temperature and humidity in real time, constructing a spatio-temporally aligned multi-modal feature matrix and environmental feature tensor; using a neural network to predict the dynamic evolution of the terrain and the trend of climate mutations, generating a scene change trend vector to dynamically adjust the parameters of the hybrid model; generating a hybrid feature vector based on feature expansion, normalization and differential processing, and synergistically mapping the air pressure target value through fuzzy logic and non-linear optimization; finally driving the bidirectional pressure regulating device to respond quickly, making the actual air pressure value stable within the preset fluctuation threshold, realizing high-precision adaptive control of the target air pressure in a complex dynamic environment, and then realizing precise attitude control of the hybrid target in this environment.

[0026] Figure 1The flowchart of an attitude control method for a hybrid target provided by an embodiment of this application is as follows Figure 1 As shown, this method includes: S11. Obtain the moving speed data of the movable entity target vehicle, the hardness level parameter data and the actual air pressure value of the hybrid target, as well as the terrain data and temperature and humidity data of the environment where the movable entity target vehicle is located.

[0027] Among them, the moving speed data is the real-time movement rate of the movable entity target vehicle under a specific terrain, the hardness level parameter data is the quantified index of the air pressure level corresponding to the current hardness of the hybrid target, the actual air pressure value is the real-time measurement value of the internal pressure sensor of the target, the terrain data is the three-dimensional terrain feature vector such as the slope and surface roughness of the position where the target vehicle is located, and the temperature and humidity data is the synchronous monitoring value of the environmental temperature and relative humidity. The hybrid target is a hybrid target located on the movable entity target vehicle and can be an inflatable target. Therefore, pressure control is required to realize the attitude control of the hybrid target.

[0028] In the embodiment of this application, first, the real-time moving speed data of the movable entity target vehicle is obtained through a sensor, and at the same time, the hardness level parameter data and the current actual air pressure value recorded by the built-in pressure sensor of the hybrid target are read. The three-dimensional terrain data of the environment where the movable entity target vehicle is located is collected by a lidar scanner, and the environmental temperature and relative humidity data are continuously monitored by a temperature and humidity sensor. All data are transmitted to the vehicle-mounted central processing unit through a bus to form a multi-dimensional data set including moving speed, hardness parameter, air pressure value, terrain features and environmental temperature and humidity.

[0029] S12. According to the terrain data and temperature and humidity data, predict the scene change trend through a pre-trained neural network model to obtain a scene change trend vector, and adjust the parameters of the target hybrid model according to the scene change trend vector. The target hybrid model is a hybrid model integrating a fuzzy logic module and a non-linear optimization module.

[0030] Among them, the scene change trend prediction can be based on the spatio-temporal correlation of terrain and temperature and humidity data to predict the evolution direction of environmental parameters within the next 10 seconds. The target hybrid model is a parallel coupling architecture including a fuzzy logic module and a non-linear optimization module. The fuzzy logic module is a rule inference system based on fuzzy set theory, which is used to process inaccurate, unstructured or input data with fuzzy boundaries. The non-linear optimization module is a mathematical modeling tool used to find the global optimal solution of the objective function under complex constraints.

[0031] In the embodiment of the present application, a deep convolutional neural network is used to extract features from terrain elevation data and temperature and humidity time-series data, and a pre-trained model is used to predict the scene change trend within the next 15 seconds. The predicted 16-dimensional scene change trend vector is input into a hybrid model including a fuzzy logic controller, where the fuzzy logic module processes the terrain slope and temperature and humidity correlation parameters through membership functions, and the non-linear optimization module uses an improved particle swarm algorithm to dynamically adjust the air pressure compensation weight coefficient, and finally generates a set of hybrid model configuration parameters including multiple key parameters.

[0032] S13. Based on the moving speed data, hardness level parameter data, terrain data, and temperature and humidity data, calculate the air pressure target value through the target hybrid model with adjusted parameters.

[0033] Among them, the moving speed data, hardness level parameter data, terrain data, and temperature and humidity data can be spliced into a hybrid input vector after being processed by feature engineering. The air pressure target value is the optimal air pressure setting value that meets the hardness requirements of the current scene.

[0034] In the embodiment of the present application, the obtained moving speed data and hardness level parameters are normalized, combined with the terrain elevation gradient value and the temperature and humidity compensation coefficient, and input into the target hybrid model with adjusted parameters. The model first calculates the basic air pressure compensation amount through a fuzzy inference system, and then performs secondary correction through a non-linear optimizer. Finally, a weighted summation formula is used, and the formula is: moving speed multiplied by the terrain resistance coefficient plus the humidity influence factor multiplied by the temperature compensation term, to generate an accurate air pressure target value, and the output result is a target air pressure command with an accuracy of ±0.5 kPa.

[0035] S14. According to the air pressure target value, control the gas injection or release rate of the two-way pressure regulating device of the hybrid target until the fluctuation amplitude of the actual air pressure value after pressure regulation is less than the preset fluctuation amplitude threshold, and the stable duration is greater than the preset duration threshold. After the control is completed, control the attitude of the hybrid target according to the actual air pressure value. The two-way pressure regulating device includes an air pump and a pressure relief valve.

[0036] Among them, the gas injection or release rate is used to adjust the torque of the drive motor of the air pump, the air pump power, etc. The fluctuation amplitude of the actual air pressure value is less than the preset fluctuation amplitude threshold, and the stable duration is greater than the preset duration threshold, which is used to reflect that the actual air pressure value after pressure regulation is stable at the air pressure target value. And there is a corresponding relationship between the air pressure target value and the attitude parameters. The attitude can be accurately controlled according to the attitude parameters corresponding to the air pressure target value.

[0037] In the embodiment of the present application, based on the difference between the target air pressure value and the actual air pressure value, a control algorithm is used to adjust the two-way pressure regulating device. Exemplarily, when the target air pressure value is higher than the actual air pressure value, the air injection rate is calculated according to the formula of target difference × 0.8 + integral of historical deviation × 0.2, and the driving air pump is modulated based on the injection rate; when the target air pressure value is lower than the actual air pressure value, the pressure relief valve is controlled to open at a proportional release rate. The air pressure fluctuation amplitude is continuously monitored. When the fluctuation value is less than ±0.3 kPa within 30 seconds and is stably maintained for 120 seconds, the system steady-state locking mechanism is triggered, and at the same time, the attitude of the hybrid target is controlled according to the actual air pressure value after the control is completed.

[0038] The following is a specific example: In the training scenario of a desert shooting range, a movable entity target vehicle maneuvers in an S shape at a speed of 8 km / h. The temperature and humidity sensor measures the ambient temperature of 42 °C and the humidity of 15%. The terrain scan finds a 5° slope sandy land ahead, and the neural network predicts that air flow disturbances will be encountered in the future. The hybrid model calculates that the target air pressure should be increased from 85 kPa to 89 kPa. The air pump rapidly pressurizes in a pulse mode. When it is detected that the air pressure reaches 88.7 kPa and the fluctuation is less than 0.25 kPa for 125 seconds, the system enters the steady-state holding mode. After the inflation is completed, the attitude of the hybrid target is controlled according to the actual air pressure value. The whole process enables the hybrid target to maintain the best attitude in a complex environment.

[0039] Optionally, in the case where the same target corresponds to different actual air pressure values, for the same attitude, such as the control of being perpendicular to the ground, the magnitudes of the required numerical value of the equipment traction force, the required numerical value of the tightness of the target mounting bracket, etc. are different. Taking being perpendicular to the ground as an example, the greater the actual air pressure value, the greater its hardness, so the required numerical value of the equipment traction force is smaller. On the contrary, the smaller the actual air pressure value, the lower its hardness, so it is more affected by the environment. To maintain verticality, a larger numerical value of the equipment traction force is required because it is accurate and reasonable to adjust the attitude of the hybrid target according to the actual air pressure value.

[0040] The following is another specific example: A movable entity target vehicle that is low-cost, movable, and easy to replace the target can meet the diverse needs in different training or test scenarios. The movable entity target vehicle mainly consists of a chassis system and an upper-mounted inflatable target system, where: (1) The chassis system is designed as follows: Steel is used as the main material for the vehicle frame, and a truss structure is adopted, which is welded by steel pipes to form a strong and lightweight vehicle frame. Reinforcing beams are designed at the front and rear parts to improve the anti-impact ability. The tires are made of wear-resistant rubber materials and are equipped with anti-slip patterns to improve the grip on different ground surfaces. The support feet are designed to be adjustable in height and are connected to the chassis by bolts to adapt to uneven ground. A simple and efficient drive system is designed, which is driven by an electric motor and can control the moving path and speed of the target vehicle through remote control or preset programs.

[0041] (2) The inflatable target system is designed as follows: A modular target mounting frame is designed, which can be quickly disassembled and different types of targets can be replaced. The target mounting frame can be connected to the chassis by means of snap fasteners, bolts or magnetic attraction, ensuring firmness and easy operation. The inflatable target is made of high-strength and wear-resistant polyvinyl chloride material, formed by heat sealing and sewing processes, and connected to the air pressure regulating device installed on the chassis through a connecting pipe, and the hardness and stability of the target can be adjusted as needed. A variety of target shapes and sizes are provided, such as launch vehicles, radar vehicles, power supply vehicles, etc., to meet different training or test requirements. In addition, the control system in this movable physical target vehicle integrates a wireless short-wave radio in the chassis, and functions such as the movement control of the target vehicle, target replacement and air pressure adjustment are realized through a remote control or a command center console. An intelligent control system is equipped, and the automatic movement of the target vehicle and target transformation are triggered through a preset program to improve the automation degree of training or testing. In terms of safety design, in the embodiment of the present application, a smooth design is adopted for the edges of the target vehicle and the target to reduce the risk of injury during accidental collisions. And an emergency stop button and an overload protection device are equipped to ensure safety during the operation process.

[0042] By executing S11~S14, the embodiment of the present application realizes the adaptive dynamic regulation of the air pressure of the hybrid target by fusing the environment prediction model and the intelligent control strategy, and further realizes the precise control of the attitude. Based on the prediction ability of the scene change trend, combined with the hybrid model of fuzzy logic and non-linear optimization, the control robustness under complex terrain and changing climate conditions is improved. The cooperative control mechanism of the two-way pressure regulating device effectively reduces the amplitude of air pressure fluctuation and shortens the system stabilization time, enabling the hybrid target to maintain the best hardness characteristics during dynamic movement and enhancing its adaptability to different training scenarios.

[0043] In a possible embodiment, S13, based on the moving speed data, hardness level parameter data, terrain data and temperature and humidity data, calculates the air pressure target value through the target hybrid model after parameter adjustment, including: Step 131, align the sampling timestamps of the moving speed data and the hardness level parameter data to generate a multi-feature matrix.

[0044] Among them, the alignment uses a dual-threshold comparator as the alignment method, and the multi-feature matrix is a spatio-temporal correlation data matrix generated by aligning the timestamps of heterogeneous sensors such as moving speed and hardness level, and includes cross-dimensional features such as speed mean and hardness change rate.

[0045] In the embodiment of the present application, first, the time synchronization module aligns the time stamps of the moving speed data collected by the moving speed sensor and the hardness level parameter data recorded by the hardness level parameter sensor. The linear interpolation algorithm is used to resample the data streams with inconsistent sampling frequencies, and the moving speed data points at intervals of 0.1 second are matched with the corresponding hardness level parameter data points according to the time axis to generate a two-dimensional multi-feature matrix containing two columns of features. The row dimension of this matrix represents the time series, and the column dimension represents different sensor data, ensuring the temporal consistency of subsequent feature fusion.

[0046] Step 132: Perform second-order polynomial basis expansion on the multi-feature matrix to generate an expanded feature matrix, perform local response normalization on the environmental feature tensor to generate a normalized tensor, and calculate the differential feature vector based on the time series of barometric pressure historical values. The environmental feature tensor is generated according to the terrain data and the temperature and humidity data.

[0047] Among them, the expanded feature matrix is a high-dimensional matrix generated by non-linearly combining the original features through second-order polynomial basis functions; local response normalization is an operation of non-linearly scaling environmental features such as terrain slope and curvature; the differential feature vector is a rate-of-change vector calculated based on the sliding window of the barometric pressure historical sequence.

[0048] In the embodiment of the present application, by calculating the square term of the speed value, the square term of the hardness value, and the product cross-term of the speed and the hardness, the original 2-column feature is expanded into an expanded feature matrix with 5 columns of features. At the same time, the terrain data and the temperature and humidity data are combined into a 4D environmental feature tensor, and the local response normalization technology is used to normalize the eigenvalue within the adjacent 3 time windows centered on each feature point to generate a normalized tensor. In addition, based on the time series of barometric pressure historical values, the first-order difference between adjacent time stamps is calculated to generate a differential feature vector.

[0049] Step 133: Perform weighted splicing on the expanded feature matrix, the normalized tensor, and the differential feature vector to form a mixed feature vector.

[0050] Among them, weighted splicing is an operation of fusing feature vectors with different dimensions in the latent space by assigning weight coefficients through the entropy weight method.

[0051] In the embodiment of the present application, first, a weight coefficient of 0.6 is assigned to the expanded feature matrix, a weight of 0.3 is assigned to the normalized tensor, and a weight of 0.1 is assigned to the differential feature vector, and a 10D mixed feature vector is generated through the feature dimension splicing formula. Among them, the terrain-related features dominate the weight in the normalized tensor, ensuring the influence of environmental factors in the final features.

[0052] Step 134: Perform non-linear mapping on the mixed feature vector through the target mixed model with adjusted parameters to obtain the barometric pressure target value.

[0053] Among them, the non-linear mapping is an operation that maps a high-dimensional feature vector to a target air pressure value through a hybrid model.

[0054] In the embodiment of the present application, the target hybrid model after parameter adjustment receives the hybrid feature vector as input and performs non-linear mapping through a three-layer fully connected neural network: the first layer uses an activation function to perform non-linear transformation on the features, the second layer generates intermediate feature weights through an S function, and the third layer uses a linear regression layer to map the weighted features to the air pressure target value. Gradient clipping is performed before the model output to prevent numerical overflow, and the finally generated air pressure target value is strictly matched with the current environmental state and the dynamic characteristics of the target.

[0055] The following is a specific example: During the training in the desert shooting range, when the movable entity target vehicle performs an S-shaped maneuver at 8 km / h, the system aligns the timing data of the moving speed sensor and the hardness sensor in real time: the speed value and the target hardness value are unified to a time granularity of 0.01 seconds through cubic spline interpolation to generate a 1200×2 multi-feature matrix. Subsequently, a second-order polynomial basis expansion is performed, adding speed squared terms, hardness squared terms, and speed-hardness cross terms to form a 1200×5 expanded feature matrix; at the same time, the environmental feature tensor constructed by a 5° slope, 42°C temperature, and 15% humidity is locally response-normalized, and a differential feature vector [1,1] is generated in combination with the historical air pressure value. After predicting the air flow disturbance, the system dynamically adjusts the feature weights: the expanded features are given a weight of 0.65 to strengthen the non-linear relationship, the normalized environmental features are given a weight of 0.3 to highlight the slope influence, and the differential features are given a weight of 0.05, and finally they are spliced into a 10-dimensional hybrid feature vector. The target hybrid model analyzes this vector through a three-layer neural network: the first layer activation captures the non-linear coupling of speed and slope, the second layer S function weighting determines that the air pressure needs to rise rapidly, and the final output corrects the target air pressure from 89 kPa to 89.8 kPa to compensate for the air flow interference. Accordingly, the air pump switches to the high-frequency pulse mode and quickly raises the air pressure from 85 kPa to 88.7 kPa within 12 seconds. When the fluctuation stabilizes within ±0.25 kPa for 125 seconds, the target still maintains the best impact feedback in the dust and sudden air flow, and the complete closed-loop verifies the effectiveness of multi-feature fusion and dynamic compensation.

[0056] By executing steps 131 to 134, the embodiment of the present application improves the accuracy and robustness of the prediction of the air pressure target value through multi-dimensional feature fusion and non-linear transformation. Timestamp alignment ensures the temporal consistency of multi-source data, polynomial expansion enhances the model's ability to capture non-linear relationships, local response normalization effectively suppresses environmental noise interference, the weighted splicing mechanism strengthens the decision weights of key features, and finally a high-precision mapping from complex features to the target air pressure is achieved through a deep neural network, enabling the target pressure regulation to have both fast response and steady-state holding characteristics.

[0057] In a possible embodiment, in step 132, perform a second-order polynomial basis expansion on the multi-feature matrix to generate an expanded feature matrix, perform local response normalization on the environmental feature tensor to generate a normalized tensor, calculate a differential feature vector based on the time series of barometric pressure history values, and the environmental feature tensor is generated according to terrain data and temperature and humidity data, including: Step a1: For each feature dimension in the multi-feature matrix, calculate the original feature values on the feature dimension, the product terms of pairwise feature dimensions, and the square terms of individual feature dimensions, and splice the original feature values, product terms, and square terms by column to generate an expanded feature matrix.

[0058] Among them, the expanded feature matrix is a high-dimensional data matrix generated by performing a second-order combination on the original features, including the original feature values, pairwise feature product terms, and single-feature square terms, and is used to capture the non-linear coupling effect between multi-dimensional parameters.

[0059] In the embodiment of the present application, first traverse each feature dimension of the multi-feature matrix, extract the original feature values for each feature dimension, calculate the product terms between pairwise feature dimensions, and the square terms of individual feature dimensions. Splice all the original feature values, product terms, and square terms by column. For example, the original 2D feature matrix is expanded to generate a 5D expanded feature matrix including the original column, product column, and square column, realizing the non-linear enhancement of the feature space.

[0060] Step a2: For each channel of the environmental feature tensor, select the sum of the squared activation values of adjacent m channels as the normalization base, weight the normalization base with a preset channel attenuation factor, divide the activation value of the current channel by the weighted normalization base to obtain a normalized value, and perform non-linear compression on the normalized value through the hyperbolic tangent function to generate a normalized tensor.

[0061] Among them, the normalized tensor is a standardized expression of the environmental features calculated by weighted sum of the squared activation values across channels, and is used to suppress the numerical offset caused by terrain mutations.

[0062] In the embodiment of the present application, for each channel of the environmental feature tensor, select the sum of the squared activation values of its adjacent m channels as the normalization base. Weight the normalization base with a preset channel attenuation factor, divide the activation value of the current channel by the weighted normalization base to obtain a preliminary normalized value. Then perform non-linear compression on the normalized value through the hyperbolic tangent function to constrain the numerical range within the interval [-1, 1], and finally generate a normalized tensor.

[0063] Step a3: Using a fixed window length, extract the barometric pressure historical value sequences within multiple windows from the time series of barometric pressure historical values. Calculate the absolute difference sequences between the barometric pressure historical values at adjacent time points within each window, perform moving average filtering on the absolute difference sequences, and obtain the filtered sequences corresponding to each window. Accumulate and sum the filtered sequences corresponding to all windows to generate a differential feature vector.

[0064] Among them, the differential feature vector is a vector that extracts the barometric pressure change trend through a sliding window and is used to characterize the short-term fluctuations caused by environmental mutations. Moving average filtering is a technique for smoothing time series data, aiming to reduce the influence of random noise on the barometric pressure difference sequence at adjacent time points, so as to extract more stable trend features.

[0065] In the embodiment of the present application, multiple sliding window sequences are intercepted from the time series of barometric pressure historical values with a fixed window length. For the barometric pressure sequence within each window, calculate the absolute difference between the barometric pressure values at adjacent time points to form an absolute difference sequence. Perform moving average filtering on this sequence to generate a filtered sequence. Finally, accumulate and sum the filtered sequences of all windows at each time point to generate a differential feature vector representing the intensity of barometric pressure change.

[0066] The following is a specific example: During training at a desert shooting range, when the movable entity target vehicle performs an S-shaped maneuver at a speed of 8.2 km / h, the system constructs a 2D multi-feature matrix for the real-time collected moving speed and target hardness. By calculating the product term of speed × hardness, the square of speed, and the square of hardness, a 5D extended feature matrix is generated. At the same time, the environmental feature tensor is composed of a 5° slope obtained from terrain scanning, a temperature of 42 °C detected by a temperature and humidity sensor, and a humidity of 15%. For the slope channel, select the adjacent temperature and humidity channels to calculate the normalization base number, and after weighting with a preset channel attenuation factor of 0.5, obtain 994.5. Divide the original slope value of 5 by this base number and apply the hyperbolic tangent function to compress it to 0.157 to generate an anti-interference normalized tensor. Based on the barometric pressure historical values, the system intercepts the sequence with a fixed window of 10 seconds, calculates the absolute difference between adjacent time points, and obtains a smooth sequence [0.67, 0.67] after 3-point moving average filtering. Accumulate to generate a differential feature vector value of 1.34 representing the barometric pressure mutation trend. The extended feature matrix, normalized slope, and differential features are input into the target hybrid model after weighted fusion, driving the air pump to quickly increase the barometric pressure from 85 kPa to 89 kPa under sand and dust disturbances, and finally stabilizing at 88.7 kPa through closed-loop control, verifying the collaborative efficiency of multi-feature extension, environmental perception normalization, and dynamic differential analysis, and ensuring that the target maintains the best posture in extreme environments.

[0067] By performing steps a1 to a3, the embodiments of the present application enhance the model's ability to capture non-linear relationships through multi-dimensional feature expansion, improve the anti-interference ability of environmental features using a channel-aware normalization mechanism, and accurately extract the air pressure change trend by combining sliding window differential analysis. The synergistic effect of feature expansion and normalization improves the robustness of target value prediction in complex environments, while the dynamic differential feature construction provides a basis for forward-looking adjustment of the system, overall achieving the coordinated optimization of air pressure control accuracy and response speed.

[0068] In a possible embodiment, step a1: Weightedly splice the extended feature matrix, the normalized tensor, and the differential feature vector to form a mixed feature vector, including: Step b1: Assign weight coefficients related to feature dimensions, weight coefficients related to the number of channels, and weight coefficients with fixed values to the extended feature matrix, the normalized tensor, and the differential feature vector respectively.

[0069] Among them, the weight coefficient related to the feature dimension is a dynamic weight calculated based on the information entropy of each dimension of the feature matrix, reflecting the importance of data distribution in different feature dimensions. The weight coefficient related to the number of channels is a channel attention vector generated by an attention mechanism according to the number of channels of the input data. The fixed-value weight coefficient is a predefined constant weight used to ensure the stability of the basic features.

[0070] Step b2: According to the weight coefficient related to the feature dimension, the weight coefficient related to the number of channels, and the weight coefficient with a fixed value, perform Hadamard product operations on the extended feature matrix, the normalized tensor, and the differential feature vector respectively to obtain the weighted extended feature matrix, the weighted normalized tensor, and the weighted differential feature vector.

[0071] Among them, the Hadamard product is an element-wise multiplication of matrices, used to fuse the weight coefficients with the original features element by element. The formula for the Hadamard product is: , where is the original extended feature matrix, is the weight coefficient matrix related to the feature dimension, is the Hadamard product operator, is the weighted extended feature matrix.

[0072] Step b3: Perform format conversions on the weighted extended feature matrix and the weighted normalized tensor respectively to obtain an extended feature vector and a normalized vector.

[0073] Among them, the format conversion is to convert a high-dimensional tensor into a vector form that matches the dimension of the target feature space.

[0074] Step b4: Concatenate the elements at the same positions in the extended feature vector, the normalized vector, and the weighted differential feature vector according to a preset concatenation order to form a mixed feature vector.

[0075] Among them, the mixed feature vector refers to a comprehensive feature vector formed by concatenating feature vectors from different sources according to a preset order. The purpose is to jointly express features with multi-dimensions and multi-physical meanings in a unified vector space, enhancing the model's global perception ability for complex scenarios. For example, if the extended feature vector contains the non-linear combination of speed and hardness, the normalized vector contains the compressed terrain parameters, and the differential feature vector contains the air pressure change rate, the mixed feature vector integrates this cross-dimensional information into a single vector for subsequent joint reasoning by the model.

[0076] The following is a specific example: In the desert shooting range test scenario, first, collect the extended feature matrix through the target vehicle sensor and assign the weight coefficient matrix related to the feature dimensions 0.7, 0.7, 0.6, 0.6; at the same time, assign the weight coefficients related to the number of channels 0.4, 0.5, 0.3, 0.3 to the normalized tensor generated by the terrain sensor, and assign the fixed numerical weight coefficient 0.2 to the differential feature vector. Secondly, perform the Hadamard product operation on the extended feature matrix: multiply the speed 9.5m / s × 0.7 = 6.65, hardness 7.8 × 0.7 = 5.46, temperature 42 × 0.6 = 25.2, humidity 12 × 0.6 = 7.2 to generate the weighted extended feature matrix 6.65, 5.46, 25.2, 7.2; perform the Hadamard product operation on the normalized tensor: multiply the slope 15° × 0.4 = 6.0, roughness 0.9 × 0.5 = 0.45, temperature 42°C × 0.3 = 12.6, humidity 12% × 0.3 = 3.6 to generate the weighted normalized tensor 6.0, 0.45, 12.6, 3.6; for the differential feature vector 3.2kPa / s × 0.2 = 0.64 to generate the weighted differential feature vector 0.64. Then, flatten the weighted extended feature matrix and the normalized tensor into vectors 6.65, 5.46, 25.2, 7.2 and 6.0, 0.45, 12.6, 3.6 respectively. Finally, concatenate the three in the order of "extended, normalized to differential" to generate the mixed feature vector 6.65, 5.46, 25.2, 7.2, 6.0, 0.45, 12.6, 3.6, 0.64. After inputting it into the model, the output target air pressure value is 96.8kPa, with an error of only 1.3kPa compared to the measured value of 98.1kPa, and the air pressure is maintained stable for 28 seconds in the sandstorm, verifying the effectiveness of the weight allocation and feature fusion mechanism.

[0077] By performing steps b1 to b4, the embodiments of the present application achieve efficient fusion of multi-source heterogeneous features by dynamically allocating feature dimensions, channel numbers, and fixed weight coefficients, combined with Hadamard product operations and format conversions. The extended feature matrix enhances the expression ability of spatial information, the normalized tensor improves the inter-channel correlation modeling, and the differential feature vector captures local detail changes. The finally generated hybrid feature vector has strong compatibility and is suitable for downstream tasks such as classification and detection, improving the model's characterization ability for complex data. In a possible embodiment, S13, the target hybrid model after parameter adjustment performs a non-linear mapping on the hybrid feature vector to obtain the air pressure target value, including: Step c1, extract a first sub-vector related to terrain and target hardness and a second sub-vector related to environment and moving speed from the hybrid feature vector.

[0078] Among them, the first sub-vector is a feature subset directly related to terrain and target hardness extracted from the hybrid feature vector, used to characterize the physical interaction effect between the machine and the terrain. The second sub-vector will extract a feature subset related to moving speed and environmental parameters, used to characterize the coupling effect of dynamic motion and environmental disturbance.

[0079] Step c2, input the first sub-vector and the second sub-vector into the fuzzy logic module and the non-linear optimization module respectively. The fuzzy logic module calculates the fuzzy inference result according to the predefined fuzzy logic rule base, combined with the weight parameters of the adjusted membership function, and converts the fuzzy inference result into a first air pressure compensation value. The non-linear optimization module calculates the air pressure reference value based on the penalty factor and iteration step size of the adjusted non-linear optimization algorithm.

[0080] Among them, the fuzzy logic rule base is a set of conditional rules based on expert experience or data-driven, used to describe the non-linear mapping relationship between input features and output actions (such as air pressure compensation values). The membership function is used to map an exact input value (such as slope = 0.12) to the membership degree (a value between 0 and 1) of a fuzzy linguistic variable (such as "steep" or "gentle"). The membership function formula is: , where is the input feature value, is the central value of the fuzzy linguistic variable, is the standard deviation, controlling the function width, is the logarithm. The fuzzy inference result is a set of fuzzy output calculated through the fuzzy rule base and membership functions, which needs to be converted into an exact value through defuzzification (such as the centroid method). The first air pressure compensation value is the air pressure adjustment amount output by the fuzzy logic module, which is used to compensate for the impact of terrain mutations or target hardness changes on the system. The penalty factor is the coefficient used to penalize violations of constraints in the nonlinear optimization algorithm, which controls the strictness of the constraints during the optimization process. The iteration step size is the amplitude of parameter update in each iteration of the nonlinear optimization algorithm, which affects the convergence speed and stability. The air pressure reference value is the preliminary air pressure target value calculated by the nonlinear optimization module based on the dynamic model and environmental constraints, without considering fuzzy factors such as terrain mutations.

[0081] Step c3: Superimpose the first air pressure compensation value and the air pressure reference value to obtain the air pressure target value.

[0082] Among them, the air pressure target value refers to the final air pressure setting value obtained by superimposing the first air pressure compensation value output by the fuzzy logic module and the air pressure reference value calculated by the nonlinear optimization module.

[0083] The following is a specific example: In the desert shooting range test, when the system detects that the target vehicle passes through a 5° sand slope at 8.2 km / h, the fuzzy logic module outputs a +3.2 kPa compensation value according to the "large slope + low humidity" rule combination. At the same time, the nonlinear optimization module calculates an 85.8 kPa reference value considering the material expansion effect caused by high temperature; during the superimposition process, the system automatically limits the compensation value not to exceed the safety threshold, and finally generates an 89.0 kPa target value, driving the air pump to complete pressurization within 12 seconds, enabling the target to maintain an operating pressure of 88.7 ± 0.2 kPa in the sandstorm environment, verifying the effectiveness of the dual-module collaborative control.

[0084] By executing steps c1~c3, the embodiments of the present application achieve specialized processing of different modules through intelligent segmentation of feature subvectors. The fuzzy logic module effectively captures the complex nonlinear relationship between terrain and hardness, and the nonlinear optimization module accurately quantifies the dynamic impact of environment and speed. The collaborative superposition of the outputs of the two modules not only maintains the reference stability but also has the ability to quickly respond to disturbances.

[0085] In a possible embodiment, S12: According to the terrain data and temperature and humidity data, predict the scene change trend through a pre-trained neural network model to obtain the scene change trend vector, and adjust the parameters of the target hybrid model, including: Step 121: Discretize the terrain data into a terrain elevation matrix according to the preset grid cells, map the temperature and humidity data into a temperature and humidity distribution tensor according to the spatial coordinates, and perform cross-modal splicing on the terrain elevation matrix and the temperature and humidity distribution tensor to generate an environmental feature tensor.

[0086] Among them, the terrain elevation matrix discretizes continuous terrain elevation data into a two-dimensional matrix according to a preset grid, and each grid cell stores the elevation value at the corresponding position. The temperature and humidity distribution tensor maps the point data of temperature and humidity sensors into a three-dimensional tensor through spatial interpolation, and the channels correspond to temperature and humidity respectively. Cross-modal splicing concatenates the terrain matrix and the temperature and humidity tensor along the channel dimension to form a unified environmental feature tensor containing terrain, temperature, and humidity.

[0087] Step 122: Input the environmental feature tensor into a pre-trained neural network model, combine the historical environmental feature tensors within the time sliding window, predict the scene change trend, and output a scene change trend vector representing the dynamic evolution of the terrain and the intensity of climate fluctuations.

[0088] Among them, the scene change trend vector is a multi-dimensional vector output by the neural network model, including components such as the intensity of terrain dynamic evolution and the intensity of climate fluctuations. The environmental feature tensor is a representation form that integrates multiple environmental parameters into a unified multi-dimensional data structure for machine learning and model processing.

[0089] Step 123: According to the terrain-related components in the scene change trend vector, adjust the weight parameters of the membership functions in the corresponding fuzzy logic rule base of the fuzzy logic module, and according to the climate-related components in the scene change trend vector, adjust the penalty factor and iteration step size of the non-linear optimization algorithm used by the non-linear optimization module.

[0090] Among them, the terrain-related components are the parameters in the trend vector that describe the terrain evolution and are used to adjust the membership functions of the fuzzy logic. The climate-related components are the parameters in the trend vector that describe the climate fluctuations and are used to adjust the penalty factor and step size of the optimization algorithm.

[0091] The following is a specific example: In a desert shooting range training scenario, the terrain is scanned by an on-vehicle lidar, and the continuous elevation data of the undulating sand dunes is discretized into grid cells of 0.5m×0.5m to generate a two-dimensional terrain elevation matrix containing 5° slope sandy land. At the same time, the real-time data collected by the temperature and humidity sensor network deployed around the shooting range is mapped into a three-dimensional temperature and humidity distribution tensor aligned with the terrain grid through the inverse distance weighted interpolation algorithm. Then, the terrain matrix and the climate tensor are cross-modally stitched along the channel dimension to form an environmental feature tensor integrating the sandy land slope and temperature and humidity dynamics. This tensor is input into a pre-trained spatio-temporal convolutional neural network, combined with the historical data within a 30-minute sliding window in the past, to predict the future 10-minute scene trend: the terrain component outputs the slope increase rate of the sand dunes caused by wind erosion, and the climate component outputs the local airflow intensity fluctuation caused by the maneuvering disturbance of the target vehicle. According to the terrain component, the weights of the membership function of "sandy land softness" in the fuzzy logic module are adjusted, and the center value of the Gaussian function is shifted from loose sandy land to semi-solidified sandy land to match the dynamic evolution of the sand dunes. At the same time, according to the climate component, the penalty factor of the non-linear optimization algorithm is increased, and the iteration step of the air pressure regulation is shortened, so that the air pump can respond quickly in a pulse mode when the air flow changes suddenly, and maintain the wind pressure resistance stability of the hybrid target after the air pressure reaches the steady-state threshold, and finally achieve the precise attitude control of the target in a complex environment.

[0092] By executing steps 121 to 123, the embodiment of the present application realizes the spatial alignment and joint modeling of terrain and climate data through the fusion of environmental feature tensors, extracts the temporal features and predicts the trends of the complex non-linear relationships of the shooting range dynamic scenes by using the spatio-temporal neural network, and enables the system to adapt to the sandy land slope evolution and air flow intensity fluctuation and optimize the air pressure regulation strategy in real time through the parameter dynamic adjustment mechanism of the fuzzy logic and the optimization algorithm.

[0093] In a possible embodiment, step 123, according to the terrain-related component in the scene change trend vector, adjust the weight parameters of the membership function in the corresponding fuzzy logic rule base of the fuzzy logic module, and according to the climate-related component in the scene change trend vector, adjust the penalty factor and iteration step of the non-linear optimization algorithm adopted by the non-linear optimization module, including: Step d1, analyze the terrain dynamic component in the scene change trend vector to extract the slope change rate and elevation fluctuation intensity, and analyze the climate fluctuation component in the scene change trend vector to extract the temperature and humidity mutation direction and wind speed correlation intensity.

[0094] Among them, the slope change rate is the change rate of the terrain slope per unit time, quantifying the dynamic evolution of the terrain steepness. The elevation fluctuation intensity is the change amplitude or fluctuation degree of the terrain elevation within a specific time period, usually expressed by the standard deviation or the change rate. The temperature and humidity mutation direction is the change trend direction of the temperature and humidity parameters in a short time, and the wind speed correlation intensity is the statistical correlation intensity between the wind speed change and other climate parameters, usually expressed by the correlation coefficient or the regression coefficient.

[0095] Step d2: According to the slope change rate and the elevation fluctuation intensity, adjust the weight parameters of the membership function of the first rule and the weight parameters of the membership function of the second rule in the fuzzy logic rule base respectively. The first rule is the rule between the slope and the hardness compensation, and the second rule is the rule between the terrain flatness and the air pressure stability.

[0096] Among them, the first rule is the association rule defining the relationship between the slope and the target hardness for air pressure compensation, and the second rule is the association rule defining the relationship between the terrain flatness and the air pressure stability. First, for the slope change rate, automatically adjust the "steep slope" judgment criterion: when the slope accelerates to become steeper, the system will relax the judgment threshold of the "steep slope" and enhance the weight of the relevant rule, making the fuzzy logic more sensitive to trigger air pressure compensation. Secondly, for the elevation fluctuation intensity, dynamically adjust the judgment range of the "terrain flatness": if the ground fluctuates violently, the system will expand the coverage interval of the "low flatness" and reduce the confidence weight of the relevant rule, so as to reduce the misjudgment caused by short-term terrain noise. Through this dynamic adaptation, the model can more accurately balance the slope risk and the ground stability in scenarios such as rainstorm collapse, improve the response speed of air pressure control, and reduce terrain-related errors.

[0097] Step d3: According to the temperature and humidity mutation direction and the wind speed correlation intensity, adjust the penalty factor and the iteration step size of the non-linear optimization algorithm respectively.

[0098] Among them, when the temperature and humidity mutation direction is detected, strengthen the application of the climate constraint by enhancing the penalty factor. For example, if the humidity mutates positively, the algorithm will proportionally increase the penalty factor, forcing the optimization process to give priority to meeting the physical limitations related to humidity and avoiding model failure due to excessive humidity; if the mutation direction is negative, appropriately reduce the penalty factor to reduce the excessive limitation on the low-temperature constraint. At the same time, dynamically adjust the iteration step size according to the wind speed correlation intensity: if the wind speed is strongly correlated with the climate parameters, reduce the step size to suppress the optimization oscillation and ensure the stable convergence of the algorithm in the wind disturbance environment; if the correlation is weak, increase the step size to accelerate the solution. For example, when the wind speed correlation intensity reaches 0.3 in the typhoon scenario, the step size is reduced by 20%, reducing the air pressure control response delay from 1.5 seconds to 1.1 seconds, and optimizing the stability standard deviation to 0.25 kPa. This adjustment mechanism ensures high precision and efficient convergence in complex scenarios by weighing the environmental mutation intensity and the algorithm robustness in real time.

[0099] The following is a specific example: During the dynamic training in the desert shooting range, when the target vehicle maneuvers in an S-shape at a speed of 8 km / h, the lidar scans the dune terrain in real time, extracts the current slope change rate and elevation fluctuation intensity. The temperature and humidity sensor synchronously detects that the environmental temperature drops suddenly from 42 °C to 38 °C, and the anemometer measures the gust correlation intensity coefficient of 0.8. According to the slope change rate, the fuzzy logic module adjusts the μ value of the Gaussian function in the interval of "slope > 5°" from 5.2 to 6.0, triggering a 15% increase in the target hardness compensation amount; based on the elevation fluctuation intensity, the weight of "high flatness" in the triangular membership function increases from 40% to 55%, driving the air pump to increase the reference air pressure from 85 kPa to 87 kPa. In response to the sudden change of temperature and humidity in the northwest direction, the penalty factor λ of the optimization algorithm increases from 1.2 to 2.5, suppressing the pressure fluctuation caused by the sudden change of temperature and humidity; combined with the wind speed correlation intensity of 0.8, the iteration step α is reduced from 0.1 to 0.0625, enabling the air pump to stabilize the air pressure at 88.7 kPa in pulse mode within 12 seconds, and controlling the fluctuation range within ±0.25 kPa. The entire mechanism finally enables the hybrid target to maintain the best posture under the sudden change of dune slope and gust disturbance.

[0100] By performing steps d1~d3, the embodiments of the present application achieve the quantitative extraction of dynamic characteristics such as sudden slope change and elevation fluctuation through the refined analysis of terrain and climate components. Combining the adaptive adjustment of the fuzzy logic rule base and the parameter dynamic optimization of the non-linear optimization algorithm, the system can respond in real time to the dune slope evolution and sudden change of temperature and humidity in complex scenarios such as desert shooting ranges, accurately adjust the target hardness compensation strategy and air pressure stability control, so as to maintain the anti-impact performance and movement trajectory stability of the target under extreme conditions such as strong wind disturbance and sand deformation, and improve the environmental adaptability and feedback authenticity of the mobile target vehicle training simulation.

[0101] Figure 2 As shown in Figure 2 the following is a schematic structural diagram of an attitude control system for a hybrid target provided by the embodiments of the present application. An acquisition module 21, configured to acquire the moving speed data of the movable entity target vehicle, the hardness level parameter data and the actual air pressure value of the hybrid target, and the terrain data and temperature and humidity data of the environment where the movable entity target vehicle is located.

[0102] An adjustment module 22, configured to predict the scene change trend through a pre-trained neural network model according to the terrain data and temperature and humidity data, obtain a scene change trend vector, and adjust the parameters of the target hybrid model according to the scene change trend vector, where the target hybrid model is a hybrid model integrating a fuzzy logic module and a non-linear optimization module.

[0103] The calculation module 23 is configured to calculate the target air pressure value through the target hybrid model with adjusted parameters based on the moving speed data, hardness level parameter data, terrain data, and temperature and humidity data.

[0104] The adjustment module 24 is configured to control the gas injection or release rate of the two-way pressure adjustment device of the hybrid target according to the target air pressure value until the fluctuation range of the actual air pressure value after pressure adjustment is less than the preset fluctuation range threshold and the stable duration is greater than the preset duration threshold. The two-way pressure adjustment device includes an air pump and a pressure relief valve.

[0105] Figure 2 The described attitude control system of the hybrid target can execute Figure 1 The attitude control method of the hybrid target described in the illustrated embodiment, the implementation principle and technical effects will not be elaborated further. For the attitude control system of the hybrid target in the above embodiment, the specific manners in which each module and unit perform operations have been described in detail in the embodiment related to the method, and will not be elaborated here.

[0106] In a possible design, Figure 2 The attitude control system of the hybrid target in the illustrated embodiment can be implemented as a computing device, such as Figure 3 shown, the computing device may include a storage component 31 and a processing component 32.

[0107] The storage component 31 stores one or more computer instructions, wherein the one or more computer instructions are called and executed by the processing component 32.

[0108] The processing component 32 is configured to: obtain the moving speed data of the movable entity target vehicle, the hardness level parameter data and the actual air pressure value of the hybrid target, and the terrain data and temperature and humidity data of the environment where the movable entity target vehicle is located. Predict the scene change trend through a pre-trained neural network model according to the terrain data and temperature and humidity data to obtain a scene change trend vector, and adjust the parameters of the target hybrid model according to the scene change trend vector. The target hybrid model is a hybrid model that integrates a fuzzy logic module and a non-linear optimization module. Calculate the target air pressure value through the target hybrid model with adjusted parameters based on the moving speed data, hardness level parameter data, terrain data, and temperature and humidity data. Control the gas injection or release rate of the two-way pressure adjustment device of the hybrid target according to the target air pressure value until the fluctuation range of the actual air pressure value after pressure adjustment is less than the preset fluctuation range threshold and the stable duration is greater than the preset duration threshold. The two-way pressure adjustment device includes an air pump and a pressure relief valve.

[0109] Among them, the processing component 32 may include one or more processors to execute computer instructions to complete all or part of the steps in the above methods. Of course, the processing component may also be implemented by one or more application-specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field programmable gate arrays (FPGAs), controllers, microcontrollers, microprocessors or other electronic components for executing the above methods.

[0110] The storage component 31 is configured to store various types of data to support the operation of the terminal. The storage component can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as random access memory (RAM), static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic memory, flash memory, magnetic disk or optical disk.

[0111] Of course, the computing device may also necessarily include other components, such as input / output interfaces, display components, communication components, etc.

[0112] The input / output interface provides an interface between the processing component and the peripheral interface module, and the above peripheral interface module may be an output device, an input device, etc.

[0113] The communication component is configured to facilitate communication between the computing device and other devices in a wired or wireless manner, etc.

[0114] Among them, the computing device may be a physical device or an elastic computing host provided by a cloud computing platform, etc. At this time, the computing device may refer to a cloud server, and the above processing component, storage component, etc. may be basic server resources leased or purchased from a cloud computing platform.

[0115] The embodiments of the present application also provide a computer storage medium storing a computer program, which when executed by a computer can implement the above-mentioned Figure 1 attitude control method of a hybrid target in the shown embodiment.

[0116] Those skilled in the art can clearly understand that for the convenience and brevity of description, the specific working processes of the above-described systems, devices, and units can refer to the corresponding processes in the foregoing method embodiments and will not be elaborated herein.

[0117] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed to multiple network units. Some or all of the modules can be selected according to actual needs to achieve the purpose of the solution of this embodiment. Those of ordinary skill in the art can understand and implement it without creative work.

[0118] From the description of the above embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus a necessary general hardware platform, and of course, it can also be implemented by hardware. Based on such an understanding, the essence of the above technical solution, or the part that contributes to the prior art, can be embodied in the form of a software product, which can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., including several instructions for causing a computer device (which can be a personal computer, server, or network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0119] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present application and are not intended to limit them. Although the present application has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions described in the foregoing embodiments or perform equivalent replacements for some of the technical features. However, these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of each embodiment of the present application.

Claims

1. A method for attitude control of a hybrid target, characterized in that, Including: Obtaining the moving speed data of the movable entity target vehicle, the hardness level parameter data of the mixed target, and the actual air pressure value, as well as the terrain data and temperature and humidity data of the environment where the movable entity target vehicle is located; According to the terrain data and temperature and humidity data, predicting the scene change trend through a pre-trained neural network model to obtain a scene change trend vector, and adjusting the parameters of the target hybrid model according to the scene change trend vector. The target hybrid model is a hybrid model integrating a fuzzy logic module and a non-linear optimization module; Based on the moving speed data, the hardness level parameter data, the terrain data and the temperature and humidity data, calculating the target air pressure value through the target hybrid model with adjusted parameters; According to the target air pressure value, controlling the gas injection or release rate of the two-way pressure regulating device of the mixed target until the fluctuation range of the actual air pressure value after pressure regulation is less than the preset fluctuation range threshold and the stable duration is greater than the preset duration threshold. After the control is completed, controlling the attitude of the mixed target according to the actual air pressure value. The two-way pressure regulating device includes an air pump and a pressure relief valve.

2. The method according to claim 1, wherein The calculating the target air pressure value through the target hybrid model with adjusted parameters based on the moving speed data, the hardness level parameter data, the terrain data and the temperature and humidity data includes: Aligning the sampling timestamps of the moving speed data and the hardness level parameter data to generate a multi-feature matrix; Performing second-order polynomial basis expansion on the multi-feature matrix to generate an expanded feature matrix, performing local response normalization on the environmental feature tensor to generate a normalized tensor, and calculating a differential feature vector based on the time series of the historical air pressure values. The environmental feature tensor is generated according to the terrain data and the temperature and humidity data; Performing weighted splicing on the expanded feature matrix, the normalized tensor and the differential feature vector to form a hybrid feature vector; Performing non-linear mapping on the hybrid feature vector through the target hybrid model with adjusted parameters to obtain the target air pressure value.

3. The method according to claim 2, wherein The performing second-order polynomial basis expansion on the multi-feature matrix to generate an expanded feature matrix, performing local response normalization on the environmental feature tensor to generate a normalized tensor, and calculating a differential feature vector based on the time series of the historical air pressure values includes: For each feature dimension in the multi-feature matrix, calculating the original feature value on the feature dimension, the product term of two feature dimensions and the square term of a single feature dimension, and splicing the original feature value, the product term and the square term by column to generate an expanded feature matrix; For each channel of the environmental feature tensor, selecting the sum of the squares of the activation values of adjacent m channels as the normalization base number, weighting the normalization base number with a preset channel attenuation factor, dividing the activation value of the current channel by the weighted normalization base number to obtain a normalized value, and performing non-linear compression on the normalized value through a hyperbolic tangent function to generate a normalized tensor; Extract sequences of barometric pressure history values within multiple windows from the time series of barometric pressure history values with a fixed window length, calculate the sequence of absolute differences between barometric pressure history values at adjacent time points within each window, perform moving average filtering on the sequence of absolute differences to obtain the filtered sequence corresponding to each window; perform cumulative summation on the filtered sequences corresponding to all windows to generate a differential feature vector.

4. The method according to claim 3, wherein The weighted concatenation of the extended feature matrix, the normalized tensor, and the differential feature vector to form a hybrid feature vector includes: Assign weight coefficients related to feature dimensions, weight coefficients related to the number of channels, and weight coefficients with fixed values to the extended feature matrix, the normalized tensor, and the differential feature vector respectively; Perform Hadamard product operations on the extended feature matrix, the normalized tensor, and the differential feature vector respectively according to the weight coefficients related to feature dimensions, the weight coefficients related to the number of channels, and the weight coefficients with fixed values to obtain the weighted extended feature matrix, the weighted normalized tensor, and the weighted differential feature vector; Perform format conversion on the weighted extended feature matrix and the weighted normalized tensor respectively to obtain an extended feature vector and a normalized vector; Concatenate the elements at the same positions in the extended feature vector, the normalized vector, and the weighted differential feature vector according to a preset concatenation order to form a hybrid feature vector.

5. The method according to any one of claims 2 to 4, characterized in that, The non - linear mapping of the hybrid feature vector by the target hybrid model with adjusted parameters to obtain the barometric pressure target value includes: Extract a first sub - vector related to terrain and target hardness and a second sub - vector related to environment and moving speed from the hybrid feature vector; Input the first sub - vector and the second sub - vector into a fuzzy logic module and a non - linear optimization module respectively. The fuzzy logic module calculates the fuzzy inference result according to a predefined fuzzy logic rule base and in combination with the weight parameters of the adjusted membership function, converts the fuzzy inference result into a first barometric pressure compensation value, and the non - linear optimization module calculates the barometric pressure reference value based on the penalty factor and iteration step size of the adjusted non - linear optimization algorithm; Superimpose the first barometric pressure compensation value and the barometric pressure reference value to obtain the barometric pressure target value.

6. The method according to claim 1, characterized in that, According to the terrain data and temperature - humidity data, predicting the scene change trend through a pre - trained neural network model to obtain a scene change trend vector, and adjusting the parameters of the target hybrid model according to the scene change trend vector includes: Discretize the terrain data into a terrain elevation matrix according to preset grid cells, map the temperature - humidity data into a temperature - humidity distribution tensor according to spatial coordinates, and perform cross - modal concatenation on the terrain elevation matrix and the temperature - humidity distribution tensor to generate an environmental feature tensor; Input the environmental feature tensor into a pre - trained neural network model, and in combination with the historical environmental feature tensors within the time sliding window, predict the scene change trend and output a scene change trend vector representing the intensity of terrain dynamic evolution and climate fluctuation. Adjust the weight parameters of the membership functions in the fuzzy logic rule base corresponding to the fuzzy logic module according to the terrain-related components in the scene change trend vector, and adjust the penalty factor and iteration step size of the non-linear optimization algorithm adopted by the non-linear optimization module according to the climate-related components in the scene change trend vector.

7. The method according to claim 6, wherein The adjusting the weight parameters of the membership functions in the fuzzy logic rule base corresponding to the fuzzy logic module according to the terrain-related components in the scene change trend vector, and adjusting the penalty factor and iteration step size of the non-linear optimization algorithm adopted by the non-linear optimization module according to the climate-related components in the scene change trend vector includes: Analyze the terrain dynamic components in the scene change trend vector to extract the slope change rate and elevation fluctuation intensity, and analyze the climate fluctuation components in the scene change trend vector to extract the temperature and humidity mutation direction and wind speed correlation intensity; According to the slope change rate and the elevation fluctuation intensity, adjust the weight parameters of the membership function of the first rule and the weight parameters of the membership function of the second rule in the fuzzy logic rule base respectively. The first rule is the rule between the slope and hardness compensation, and the second rule is the rule between the terrain flatness and air pressure stability; According to the temperature and humidity mutation direction and the wind speed correlation intensity, adjust the penalty factor and iteration step size of the non-linear optimization algorithm respectively.

8. An attitude control system for a hybrid target, characterized in that, Includes: An acquisition module for acquiring the moving speed data of the movable entity target vehicle, the hardness level parameter data of the hybrid target, and the actual air pressure value, as well as the terrain data and temperature and humidity data of the environment where the movable entity target vehicle is located; An adjustment module for predicting the scene change trend through a pre-trained neural network model according to the terrain data and temperature and humidity data to obtain a scene change trend vector, and adjusting the parameters of the target hybrid model according to the scene change trend vector. The target hybrid model is a hybrid model integrating a fuzzy logic module and a non-linear optimization module; A calculation module for calculating the target air pressure value through the target hybrid model with adjusted parameters based on the moving speed data, the hardness level parameter data, the terrain data, and the temperature and humidity data; An adjustment module for controlling the gas injection or release rate of the two-way pressure adjustment device of the hybrid target according to the target air pressure value until the fluctuation range of the actual air pressure value after pressure adjustment is less than a preset fluctuation range threshold and the stable duration is greater than a preset duration threshold. After the control is completed, control the attitude of the hybrid target according to the actual air pressure value. The two-way pressure adjustment device includes an air pump and a pressure relief valve.

9. A computing device, characterized in that, Includes a processing component and a storage component; the storage component stores one or more computer instructions; the one or more computer instructions are used to be called and executed by the processing component to implement a method for controlling the attitude of a hybrid target as described in any one of claims 1 to 7.

10. A computer storage medium, characterized in that, Stores a computer program, and when the computer program is executed by a computer, it implements a method for controlling the attitude of a hybrid target as described in any one of claims 1 to 7.

Images