Unmanned aerial vehicle aerodynamic deceleration control prediction method, electronic equipment and storage medium

Abstract

The invention discloses an unmanned aerial vehicle aerodynamic deceleration control prediction method, electronic equipment and a storage medium, and belongs to the technical field of unmanned aerial vehicle flight management. In order to improve the reliability of an unmanned aerial vehicle flight deceleration control strategy, the method comprises the steps of constructing a hidden layer of an unmanned aerial vehicle aerodynamic deceleration control prediction model, and adopting a three-layer architecture of a dynamic perception layer, an aerodynamic characteristic layer and a control decision layer; and constructing output indexes of the model, wherein the output indexes comprise a flight state prediction index and a control effect evaluation index. And constructing a composite loss function comprising a speed loss term function, a flight characteristic loss term function and a stability loss term function. Constructing a training set, a verification set and a test set on the basis of input characteristic parameter data of the unmanned aerial vehicle aerodynamic deceleration control prediction model after normalization processing, performing parameter setting on the constructed unmanned aerial vehicle aerodynamic deceleration control prediction model, then performing training by using the obtained training set, performing verification by using the verification set, and performing prediction on the unmanned aerial vehicle aerodynamic deceleration control prediction model. And testing and predicting by using the test set.

Description

Technical Field

[0001] The present invention belongs to the technical field of unmanned aerial vehicle (UAV) flight management, and in particular relates to a UAV aerodynamic deceleration control prediction method, electronic equipment and storage medium. Background Art

[0002] Aerodynamic deceleration control for unmanned aerial vehicles (UAVs) is a key technology for achieving safe and efficient landings. By rationally utilizing aerodynamic characteristics and coordinating the control of devices such as ailerons, rudders, elevators, airbrakes, and parachutes, UAVs can achieve a smooth transition from high-speed to low-speed flight while maintaining a stable attitude. This is crucial for improving UAV recovery success rates, protecting onboard equipment, extending service life, and ensuring mission integrity. However, while currently widely used deep learning models can establish a mapping relationship between input parameters and outputs, they often treat the entire control process as a black box, lacking a physical interpretation of the intermediate processes. Such models cannot effectively describe the variations in key physical quantities such as aerodynamic forces and aerodynamic characteristics, nor can they characterize the specific interaction mechanisms between control parameters and flight conditions. This lack of interpretability makes it difficult to ensure the reliability and robustness of the models, especially in complex and changing flight environments, which can lead to unpredictable control errors.

[0003] Traditional deep learning models, when addressing the aerodynamic deceleration control of drones, often overlook the dynamic changes in aerodynamic characteristics as they evolve in flight. These models rely solely on data-driven models and fail to accurately capture the mechanisms by which environmental factors such as air density, temperature, and wind speed influence aerodynamic characteristics. Furthermore, they struggle to capture the physical relationship between key parameters like Mach number and dynamic pressure and deceleration performance. This neglect of physical processes can lead to irrational control strategies in practical applications, and even produce outputs that violate fundamental aerodynamic principles.

[0004] Accurately predicting flight parameters such as airspeed, angle of attack, and altitude is crucial for UAV deceleration control. These parameters are directly related to flight safety and mission effectiveness, especially during the critical transition from high-speed to low-speed flight. However, existing models often lack an accurate understanding of the changing trends of these parameters and fail to fully account for the coupling effects of various control surfaces (such as ailerons, rudder, and elevator), as well as their interactions with aerodynamic characteristics. This severely limits the accuracy and reliability of deceleration control.

[0005] Parameters such as deceleration rate, aerodynamic drag coefficient, and lift-to-drag ratio are crucial for achieving precise deceleration control. These metrics not only reflect the effectiveness of the deceleration process but also directly impact energy management efficiency and flight safety. However, existing models often predict these metrics separately, ignoring their physical interconnectedness. This results in a lack of systematicity and coordination in the predictions. This fragmented prediction approach cannot guarantee the overall optimality of the control strategy and is difficult to meet the needs of practical engineering applications. Summary of the Invention

[0006] The problem to be solved by the present invention is to increase the reliability of the UAV flight deceleration control strategy, and propose a UAV aerodynamic deceleration control prediction method, electronic equipment and storage medium.

[0007] To achieve the above object, the present invention is implemented through the following technical solutions:

[0008] A method for predicting aerodynamic deceleration control of a UAV comprises the following steps:

[0009] S1. Collect and normalize the input characteristic parameters of the UAV aerodynamic deceleration control prediction model, including flight state parameters, control input parameters, and environmental parameters;

[0010] S2. Construct a hidden layer for a UAV aerodynamic deceleration control prediction model using a three-layer architecture consisting of a dynamics perception layer, an aerodynamic characteristics layer, and a control decision layer.

[0011] S3. Output indicators of the hidden layer of the UAV aerodynamic deceleration control prediction model obtained in step S2, including flight state prediction indicators and control effect evaluation indicators; flight state prediction indicators include predicted airspeed Y1, predicted angle of attack Y2, predicted altitude Y3, predicted Mach number Y4, and predicted dynamic pressure Y5; control effect evaluation indicators include deceleration rate Y6, aerodynamic drag coefficient Y7, lift-to-drag ratio Y8, energy loss rate Y9, and stability index Y 10 ;Establish the mapping relationship between the model hidden layer and output indicators;

[0012] S4. Construct a loss function for the UAV aerodynamic deceleration control prediction model that is a composite of the base loss function, the speed loss term, the flight characteristics loss term, and the stability loss term.

[0013] S5. Based on the normalized input characteristic parameter data of the UAV aerodynamic deceleration control prediction model obtained in step S1 and the normalized output index data of the UAV aerodynamic deceleration control prediction model obtained in step S3, a training set, a validation set, and a test set are constructed, and parameters are set for the constructed UAV aerodynamic deceleration control prediction model. Then, the UAV aerodynamic deceleration control prediction model is trained using the obtained training set, verified using the validation set, and tested using the verified UAV aerodynamic deceleration control prediction model.

[0014] Furthermore, the input characteristic parameters of the UAV aerodynamic deceleration control prediction model in step S1 are as follows:

[0015] Flight status parameters include airspeed X1, angle of attack X2, sideslip angle X3, altitude X4, Mach number X5, dynamic pressure X6, roll angle X7 and pitch angle X8;

[0016] The control input parameters include aileron deflection angle X9, rudder deflection angle X 10 , elevator deflection angle X 11 , throttle opening X 12 , air brake plate deflection angle X 13 and the deployment area of the parachute X 14 ;

[0017] Environmental parameters include air density X 15 , temperature X 16 , wind speed X 17 、wind direction X 18 , atmospheric pressure X 19 and humidity X 20 .

[0018] Furthermore, the specific implementation method of step S2 includes the following steps:

[0019] S2.1. Based on the consideration of airspeed and altitude among flight state parameters, and the impact of environmental parameters on dynamic characteristics, a dynamic perception layer is constructed. The expression is:

[0020]

[0021] in, is the output of the dynamic perception layer, is the activation function, is the weight coefficient of the input characteristic parameters of the aerodynamic deceleration control prediction model of the i-th UAV, X i is the input characteristic parameter of the aerodynamic deceleration control prediction model of the i-th UAV, is the dynamic characteristic coefficient, is a high impact factor, is the characteristic time scale factor, which characterizes the characteristic response speed of the UAV during deceleration; is the characteristic height scale factor, which represents the sensitivity of the UAV to altitude changes; is the bias term of the dynamic perception layer, i=1~20;

[0022] Dynamic characteristic coefficient The expression reflecting the dynamic characteristics of the UAV during deceleration is:

[0023]

[0024] in, For the quality of the drone, is the drag coefficient;

[0025] High impact factor The relationship between altitude and atmospheric pressure is expressed as:

[0026]

[0027] in, is the reference height, is standard atmospheric pressure;

[0028] S2.2. Based on the dynamic characteristics, the effects of angle of attack and dynamic pressure on aerodynamic forces are considered, and the Mach number effect is considered to construct the aerodynamic characteristic layer. The expression is:

[0029]

[0030] in, Output for the aerodynamic characteristics layer; is the bias term of the aerodynamic characteristic layer, is the lift characteristic coefficient; is the compressibility effect factor, is the characteristic angle of attack, which characterizes the response characteristics of the system to attitude changes during deceleration; is the characteristic dynamic pressure, which represents the adaptability of the UAV to aerodynamic changes;

[0031] lift characteristic coefficient The aerodynamic characteristics are reflected by the ratio of dynamic pressure, parachute deployment area and gravity, and the expression is:

[0032]

[0033] Where g is the acceleration due to gravity;

[0034] Compressibility effect factor Considering the influence of air density and Mach number on the compressibility effect, the expression is:

[0035] ;

[0036] S2.3. Construct a control decision layer. By establishing a direct information channel from the dynamics perception layer to the control decision layer, a rapid mapping of the initial dynamic state to the control decision is achieved. The expression is:

[0037]

[0038] in, To control the output of the decision-making layer; is the residual connection coefficient, determined by expert experience; is the control efficiency coefficient; is the stability influencing factor, To control the bias term of the decision layer; is the characteristic force, which represents the impact of the change of control force on the UAV; is the characteristic rudder deflection angle, which represents the effect of rudder surface deflection on the UAV;

[0039] Control efficiency coefficient The control efficiency is reflected by considering the relationship between dynamic pressure and characteristic length. The product of dynamic pressure and the deployment area of the parachute represents the aerodynamic control force. The expression is:

[0040]

[0041] in, is the characteristic length, is the mean aerodynamic chord length;

[0042] Stability influencing factors The stability evaluation index is constructed by Mach number, and the expression is:

[0043] .

[0044] Furthermore, the specific implementation method of step S3 is to perform a weighted combination of the outputs of the three hidden layers constructed in step S2, and introduce a nonlinear activation function and a characteristic scale factor of the corresponding indicator to achieve accurate prediction of the flight state prediction indicator and the control effect evaluation indicator. The nonlinear activation function introduced by the flight state prediction indicator is the hyperbolic tangent function, and the nonlinear activation function introduced by the control effect evaluation indicator is the sigmoid function.

[0045] Furthermore, the specific implementation method of step S4 includes the following steps:

[0046] S4.1. Design a basic loss function and use the mean square error to calculate all output indicators Y1~Y 10The mean square error between the predicted value and the measured value is the basic loss ;

[0047] S4.2. Design a speed loss function to quantify the deceleration effect by evaluating the difference between the predicted and actual airspeeds and the corresponding state transition efficiency. Taking into account the effects of air density and the deployment area of the parachute, the resulting expression is:

[0048]

[0049] in, is the weight coefficient of speed loss term; is the state conversion efficiency weight coefficient; 、 Characterize the kinetic energy levels of the predicted state and the current state respectively; 、 characterize the work levels of the predicted and actual resistance, respectively;

[0050] The weight coefficient of the speed loss term reflects the change of the deceleration effect with the flight conditions. The air density predicted speed square term represents the change characteristics of kinetic energy. At the same time, considering the air density, the deployment area of the parachute, and the mass factors, the expression is obtained as follows:

[0051]

[0052] The state conversion efficiency weight coefficient evaluates the energy conversion efficiency of the deceleration process. Considering the inertial characteristics of the UAV, the characteristics of the deceleration medium, the current motion state, and the characteristics of the deceleration device, the expression of the state conversion efficiency weight coefficient is obtained as follows:

[0053] ;

[0054] S4.3. Design a flight performance loss function. Consider the impact of the Mach number effect on aerodynamic characteristics, evaluate the accuracy of lift, drag, and torque predictions, and construct a comprehensive evaluation index of flight quality by introducing characteristic length and aerodynamic parameters. The expression for the flight performance loss function is:

[0055]

[0056] in, is the Mach number effect coefficient, is the moment weight coefficient, 、 They represent the predicted aerodynamic intensity level and the actual aerodynamic intensity level respectively; 、 Respectively characterize the predicted torque level and the actual torque level;

[0057] The Mach number effect coefficient reflects the changes in aerodynamic characteristics during the flight speed change through the Mach number, and directly reflects the intensity of the Mach number effect. The expression is:

[0058] ;

[0059] The moment weight coefficient is obtained by considering the influence of environmental characteristics, parachute area and characteristic length. The characteristic length evaluates the influence of moment balance and the square of the predicted velocity reflects the dynamic pressure effect. The expression is:

[0060] ;

[0061] S4.4. Design a stability loss function. By evaluating altitude change, static stability, and dynamic stability, a complete stability evaluation system is constructed. The expression of the stability loss function is:

[0062]

[0063] in, is the height influence coefficient, is the stability weight coefficient, and They represent the mechanical energy level of the predicted state and the mechanical energy level of the current state respectively; and denote the predicted stability moment level and the actual stability moment level, is the static stability coefficient; is the characteristic energy factor; is the characteristic moment factor;

[0064] The height influence coefficient represents the influence of height, and the expression is:

[0065]

[0066] The stability weight coefficient evaluates the influence of stability, and the square of the predicted velocity reflects the influence of dynamic pressure. At the same time, the influence of environmental characteristics, parachute area, average aerodynamic chord length and characteristic length are considered. The expression is:

[0067] ;

[0068] S4.5. Construct a UAV aerodynamic deceleration control prediction model using a composite loss function consisting of speed loss, flight characteristics loss, and stability loss. The loss function is expressed as:

[0069]

[0070] in, 、 、 、 They are 、 、 、 The weight coefficient of .

[0071] Furthermore, in step S5, the input characteristic parameter data of the normalized UAV aerodynamic deceleration control prediction model and the output index data of the normalized UAV aerodynamic deceleration control prediction model obtained in step S3 are divided into a training set, a validation set, and a test set in a ratio of 8:1:1.

[0072] Furthermore, in step S5, the model training adopts the back propagation algorithm. During the training process, the model calculates the predicted value through forward propagation, substitutes the predicted value and the true value into the loss function to calculate the loss, and then updates the network parameters through back propagation.

[0073] Furthermore, in step S5, to prevent overfitting, an early stopping strategy is adopted, and training is stopped when the validation set loss does not decrease for 10 consecutive epochs. After the model training is completed, the relationship between input and output is established, and the output result is restored to the actual physical quantity through denormalization.

[0074] An electronic device includes a memory and a processor, wherein the memory stores a computer program, and the processor implements the steps of a method for predicting aerodynamic deceleration control of a drone when executing the computer program.

[0075] A computer-readable storage medium stores a computer program, which, when executed by a processor, implements the aforementioned method for predicting aerodynamic deceleration control of a UAV.

[0076] Beneficial effects of the present invention:

[0077] The method for predicting drone aerodynamic deceleration control described in this paper establishes a deep learning prediction model based on physical mechanisms. This model accurately reflects the physical characteristics of the drone's aerodynamic deceleration process and accurately predicts key state parameters and performance indicators. This model organically integrates dynamic perception, aerodynamic characteristics, and control decisions to construct a multi-level prediction system, while ensuring the model's good interpretability and generalization capabilities. Through the design of a reasonable loss function, the model accurately captures the coupling relationship between various physical parameters, ensuring that the prediction results conform to the basic principles of aerodynamics.

[0078] The proposed method for predicting aerodynamic deceleration control for unmanned aerial vehicles (UAVs) utilizes a three-layer architecture (dynamic perception layer, aerodynamic characteristics layer, and control decision layer) to provide an in-depth description of the physical mechanisms of the deceleration process. Multi-dimensional and multi-level feature selection ensures the model's comprehensive perception of flight status and effective response to environmental changes. A carefully designed loss function system enables a comprehensive assessment of velocity variation, flight characteristics, and stability. This method not only improves prediction accuracy but also enhances the model's interpretability and practicality, providing a reliable basis for optimizing UAV deceleration control strategies. BRIEF DESCRIPTION OF THE DRAWINGS

[0079] Figure 1 This is a flow chart of a method for predicting aerodynamic deceleration control of a UAV according to the present invention;

[0080] Figure 2 This is a loss curve diagram of the training process of the UAV aerodynamic deceleration control prediction model of the present invention;

[0081] Figure 3 This is a comparison chart of the predicted value and the actual value of the stability index of the UAV aerodynamic deceleration control prediction model of the present invention. DETAILED DESCRIPTION

[0082] In order to make the objectives, technical solutions, and advantages of the present invention more clearly understood, the present invention is further described in detail below with reference to the accompanying drawings and specific embodiments. It should be understood that the specific embodiments described herein are only intended to explain the present invention and are not intended to limit the present invention. That is, the specific embodiments described herein are only some embodiments of the present invention, not all embodiments. Generally, the components of the specific embodiments of the present invention described and illustrated in the drawings herein can be arranged and designed in various different configurations, and the present invention can also have other embodiments.

[0083] Therefore, the following detailed description of the specific embodiments of the present invention provided in the accompanying drawings is not intended to limit the scope of the invention as claimed, but is merely representative of selected specific embodiments of the present invention. All other specific embodiments obtained by those skilled in the art based on the specific embodiments of the present invention without making any creative efforts shall fall within the scope of protection of the present invention.

[0084] In order to further understand the content, features and effects of the present invention, the following specific embodiments are given as examples, and the attached Figure 1 -Attached Figure 3 The detailed instructions are as follows:

[0085] Example 1:

[0086] A method for predicting aerodynamic deceleration control of a UAV comprises the following steps:

[0087] S1. Collect and normalize the input characteristic parameters of the UAV aerodynamic deceleration control prediction model, including flight state parameters, control input parameters, and environmental parameters;

[0088] Furthermore, the input characteristic parameters of the UAV aerodynamic deceleration control prediction model in step S1 are as follows:

[0089] Flight status parameters include airspeed X1, angle of attack X2, sideslip angle X3, altitude X4, Mach number X5, dynamic pressure X6, roll angle X7 and pitch angle X8;

[0090] Furthermore, airspeed X1, angle of attack X2, sideslip angle X3, and altitude X4 directly reflect the instantaneous motion state of the UAV and are the basic inputs for deceleration control. Mach number X5 and dynamic pressure X6 reflect aerodynamic characteristics and are crucial for evaluating transonic effects and structural loads. Roll angle X7 and pitch angle X8 ensure attitude stability during deceleration.

[0091] The control input parameters include aileron deflection angle X9, rudder deflection angle X 10 , elevator deflection angle X 11 , throttle opening X 12 , air brake plate deflection angle X 13 and the deployment area of the parachute X 14 ;These parameters constitute the active control variables during deceleration;

[0092] Environmental parameters include air density X 15 , temperature X 16 , wind speed X 17 、wind direction X 18 , atmospheric pressure X 19 and humidity X 20 Real-time monitoring of these parameters is crucial for the development of adaptive control strategies.

[0093] Furthermore, this multi-dimensional, multi-level feature selection ensures that the model can fully perceive the flight state, accurately control the actuators, and effectively respond to environmental changes, thereby achieving safe and efficient deceleration control. Analysis of the above parameters shows that during the training of deep learning models, different input parameters have different physical dimensions and numerical ranges. For example, the airspeed may be hundreds of meters per second, while the angle of attack varies only within a few degrees. This numerical difference will lead to uneven gradient updates during model training, affecting the model's convergence speed and accuracy. Therefore, it is necessary to normalize all input and output parameters and map them to the [0,1] interval to make different features numerically comparable, preventing certain features from dominating the model training process due to their large values. This also helps to reduce rounding errors in numerical calculations and improve the numerical stability of the model.

[0094] Specifically, the maximum and minimum value ranges of each parameter are obtained based on historical flight data statistics, and the maximum and minimum value normalization method is used for processing: among the flight status parameters, the airspeed adopts the ratio of the actual value to the historical maximum airspeed, the angle of attack and sideslip angle are mapped based on the historical data statistical range, the altitude adopts the ratio of the actual altitude to the historical maximum flight altitude, and the Mach number and dynamic pressure adopt the ratio of the actual value to the historical maximum value respectively; among the attitude angle parameters, the roll angle and pitch angle are mapped based on the historical attitude data range; the control surface deflection angle (including ailerons, rudder, elevator and brake pads) adopts the ratio of the actual deflection angle to its physical limit range; the throttle opening is normalized based on the actual control range, and the deceleration parachute deployment area adopts the ratio of the actual deployment area to the maximum design area; in terms of environmental parameters, the air density and atmospheric pressure adopt the proportion of the actual value within the historical data statistical range, the temperature is mapped based on the historical meteorological data range, the wind speed uses the ratio of the actual wind speed to the historical maximum wind speed, the wind direction angle is mapped based on the full circle angle, and the humidity is normalized based on the actual percentage range.

[0095] Table 1 shows some input data of the prediction model of UAV aerodynamic deceleration control;

[0096]

[0097] In addition, it should be noted that, in order to ensure the consistency of the physical parameters in the model, all physical quantities involved in this embodiment are first normalized and then calculated. This normalization based on historical data statistics not only ensures the numerical stability and convergence efficiency of deep learning model training, but also enables the model to better capture the inherent correlation between different physical quantities and improve prediction accuracy. During the use of the model, both input data and output data use normalized data.

[0098] S2. Construct a hidden layer for a UAV aerodynamic deceleration control prediction model using a three-layer architecture consisting of a dynamics perception layer, an aerodynamic characteristics layer, and a control decision layer.

[0099] The dynamics perception layer primarily processes basic kinematics and dynamics. The aerodynamic characteristics layer focuses on processing aerodynamic effects. The control decision layer integrates information from the first two layers, focusing on the impact of control inputs on the system. This hierarchical structure not only conforms to the inherent laws of physical phenomena but also progressively addresses all aspects, from basic physics to complex control, ensuring model interpretability and control accuracy. Each layer has a clear physical meaning and functional definition, significantly improving model training efficiency and generalization capabilities.

[0100] Furthermore, the specific implementation method of step S2 includes the following steps:

[0101] S2.1. Based on the consideration of airspeed and altitude among flight state parameters, and the impact of environmental parameters on dynamic characteristics, a dynamic perception layer is constructed. The expression is:

[0102]

[0103] in, is the output of the dynamic perception layer, is the activation function, is the weight coefficient of the input characteristic parameters of the aerodynamic deceleration control prediction model of the i-th UAV, X i is the input characteristic parameter of the aerodynamic deceleration control prediction model of the i-th UAV, is the dynamic characteristic coefficient, is a high impact factor, is the characteristic time scale factor, which characterizes the characteristic response speed of the UAV during deceleration; is the characteristic height scale factor, which represents the sensitivity of the UAV to altitude changes; is the bias term of the dynamic perception layer, i=1~20;

[0104] Dynamic characteristic coefficient The expression reflecting the dynamic characteristics of the UAV during deceleration is:

[0105]

[0106] in, For the quality of the drone, is the drag coefficient; air density affects the size of aerodynamics, the square of airspeed reflects the dynamic pressure effect, the deployment area of the parachute represents the effective area, and the drag coefficient represents the aerodynamic characteristics, while considering the effect of the angle of attack on the dynamic characteristics;

[0107] High impact factor The relationship between altitude and atmospheric pressure is expressed as:

[0108]

[0109] in, is the reference height, is standard atmospheric pressure;

[0110] High impact factor This reflects the relationship between altitude and atmospheric pressure. This takes into account the fundamental laws of atmospheric physics: atmospheric density decays exponentially with increasing altitude, a characteristic that directly impacts the aerodynamic characteristics of the aircraft. Furthermore, the ratio of atmospheric pressure during actual flight to standard atmospheric pressure directly reflects the atmospheric environment at the current flight altitude, significantly impacting the aircraft's deceleration performance.

[0111] S2.2. Based on the dynamic characteristics, the effects of angle of attack and dynamic pressure on aerodynamic forces are considered, and the Mach number effect is considered to construct the aerodynamic characteristic layer. The expression is:

[0112]

[0113] in, Output for the aerodynamic characteristics layer; is the bias term of the aerodynamic characteristic layer, is the lift characteristic coefficient; is the compressibility effect factor, is the characteristic angle of attack, which characterizes the response characteristics of the system to attitude changes during deceleration; is the characteristic dynamic pressure, which represents the adaptability of the UAV to aerodynamic changes;

[0114] lift characteristic coefficient The aerodynamic characteristics are reflected by the ratio of dynamic pressure, parachute deployment area and gravity, and the expression is:

[0115]

[0116] Where g is the acceleration due to gravity;

[0117] Compressibility effect factor Considering the influence of air density and Mach number on the compressibility effect, the expression is:

[0118] ;

[0119] S2.3. Construct a control decision layer. By establishing a direct information channel from the dynamics perception layer to the control decision layer, a rapid mapping of the initial dynamic state to the control decision is achieved. The expression is:

[0120]

[0121] in, To control the output of the decision-making layer; is the residual connection coefficient, determined by expert experience; is the control efficiency coefficient; is the stability influencing factor, To control the bias term of the decision layer; is the characteristic force, which represents the impact of the change of control force on the UAV; is the characteristic rudder deflection angle, which represents the effect of rudder surface deflection on the UAV;

[0122] Control efficiency coefficient The control efficiency is reflected by considering the relationship between dynamic pressure and characteristic length. The product of dynamic pressure and the deployment area of the parachute represents the aerodynamic control force. The expression is:

[0123]

[0124] in, is the characteristic length, is the mean aerodynamic chord length;

[0125] Stability influencing factors The stability evaluation index is constructed by Mach number. When the Mach number is close to 1, a more conservative control strategy should be required, while when the Mach number is small, a more aggressive control behavior should be allowed. The expression is:

[0126] .

[0127] S3. Output indicators of the hidden layer of the UAV aerodynamic deceleration control prediction model obtained in step S2, including flight state prediction indicators and control effect evaluation indicators; flight state prediction indicators include predicted airspeed Y1, predicted angle of attack Y2, predicted altitude Y3, predicted Mach number Y4, and predicted dynamic pressure Y5; control effect evaluation indicators include deceleration rate Y6, aerodynamic drag coefficient Y7, lift-to-drag ratio Y8, energy loss rate Y9, and stability index Y 10 ;Establish the mapping relationship between the model hidden layer and output indicators;

[0128] Furthermore, the specific implementation method of step S3 is to perform a weighted combination of the outputs of the three hidden layers constructed in step S2, and introduce a nonlinear activation function and a characteristic scale factor of the corresponding indicator to achieve accurate prediction of the flight state prediction indicator and the control effect evaluation indicator. The nonlinear activation function introduced by the flight state prediction indicator is the hyperbolic tangent function, and the nonlinear activation function introduced by the control effect evaluation indicator is the sigmoid function.

[0129] Furthermore, for the output parameters, the predicted flight state quantities are normalized using the same method as the input parameters, and the performance indicators are mapped to the range [0, 1] based on the variation range obtained from historical data statistics. Based on the relationship between the outputs of the hidden layers (dynamic perception layer, aerodynamic characteristics layer, and control decision layer) and the ultimate control objectives, a mapping relationship between the output parameters and the hidden layers is established. Considering the different characteristics of flight state prediction (Y1-Y5) and control effect evaluation (Y6-Y10) during deceleration control, the outputs of the three hidden layers are weighted and combined, and a nonlinear activation function is introduced to achieve accurate prediction of different types of output parameters.

[0130] The dynamics perception layer provides basic state information, the aerodynamic characteristics layer provides nonlinear aerodynamic characteristics, and the control decision layer provides the UAV system response characteristics. These three layers of information are integrated through weight coefficients. In the specific implementation, for each output indicator, a weighted combination of information from the three hidden layers is used, and a nonlinear activation function is introduced for output mapping to ensure the physical rationality of the prediction and evaluation results.

[0131] Flight state prediction focuses on the prediction of key physical quantities such as airspeed Y1 (reflecting the direct effect of deceleration), predicted angle of attack Y2 (monitoring stall risk), predicted altitude Y3 (ensuring safety margin), predicted Mach number Y4, and predicted dynamic pressure Y5 (evaluating aerodynamic characteristics);

[0132] The expression of flight status prediction index is:

[0133]

[0134] in, 、 、 They are the weight coefficients corresponding to the dynamic perception layer, aerodynamic characteristics layer, and control decision layer in the flight state prediction process; is the characteristic scale factor of each indicator, determined by expert experience, i=1~5; tanh is the hyperbolic tangent activation function;

[0135] The control effect evaluation focuses on the deceleration rate Y6 (quantification of control effect), aerodynamic drag coefficient Y7 (evaluation of deceleration performance), lift-to-drag ratio Y8 (measurement of efficiency), energy loss rate Y9 (optimization of energy consumption) and stability index Y 10 (Ensure safety) and other performance indicators. By using nonlinear activation functions to process the weighted combination results, it not only ensures the physical rationality of the output, but also achieves accurate prediction and evaluation of different types of indicators;

[0136] The expression of the control effect evaluation index is:

[0137]

[0138] in, 、 、 They are the weight coefficients corresponding to the dynamic perception layer, aerodynamic characteristics layer, and control decision layer in the control effect evaluation process; is the characteristic scale factor of each indicator, determined by expert experience, i=6~10; is the activation function.

[0139] S4. Construct a loss function for the UAV aerodynamic deceleration control prediction model that is a composite of the base loss function, the speed loss term, the flight characteristics loss term, and the stability loss term.

[0140] Furthermore, the specific implementation method of step S4 includes the following steps:

[0141] S4.1. Design a basic loss function and use the mean square error to calculate all output indicators Y1~Y 10 The mean square error between the predicted value and the measured value is the basic loss ;

[0142] S4.2. Design a speed loss function to quantify the deceleration effect by evaluating the difference between the predicted and actual airspeeds and the corresponding state transition efficiency. Taking into account the effects of air density and the deployment area of the parachute, the resulting expression is:

[0143]

[0144] in, is the weight coefficient of speed loss term; is the state conversion efficiency weight coefficient; 、 Characterize the kinetic energy levels of the predicted state and the current state respectively; 、 Characterize the work level of predicted resistance and actual resistance respectively;

[0145] The weight coefficient of the speed loss term reflects the change of the deceleration effect with the flight conditions. The air density predicted speed square term represents the change characteristics of kinetic energy. At the same time, considering the air density, the deployment area of the parachute, and the mass factors, the expression is obtained as follows:

[0146]

[0147] The state conversion efficiency weight coefficient evaluates the energy conversion efficiency of the deceleration process. Considering the inertial characteristics of the UAV, the characteristics of the deceleration medium, the current motion state, and the characteristics of the deceleration device, the expression of the state conversion efficiency weight coefficient is obtained as follows:

[0148] ;

[0149] S4.3. Design a flight performance loss function. Consider the impact of the Mach number effect on aerodynamic characteristics, evaluate the accuracy of lift, drag, and torque predictions, and construct a comprehensive evaluation index of flight quality by introducing characteristic length and aerodynamic parameters. The expression for the flight performance loss function is:

[0150]

[0151] in, is the Mach number effect coefficient, is the moment weight coefficient, 、 They represent the predicted aerodynamic intensity level and the actual aerodynamic intensity level respectively; 、 Respectively characterize the predicted torque level and the actual torque level;

[0152] The Mach number effect coefficient reflects the changes in aerodynamic characteristics during the flight speed change through the Mach number, and directly reflects the intensity of the Mach number effect. The expression is:

[0153] ;

[0154] The moment weight coefficient is obtained by considering the influence of environmental characteristics, parachute area and characteristic length. The characteristic length evaluates the influence of moment balance and the square of the predicted velocity reflects the dynamic pressure effect. The expression is:

[0155] ;

[0156] S4.4. Design a stability loss function. By evaluating altitude change, static stability, and dynamic stability, a complete stability evaluation system is constructed. The expression of the stability loss function is:

[0157]

[0158] in, is the height influence coefficient, is the stability weight coefficient, and They represent the mechanical energy level of the predicted state and the mechanical energy level of the current state respectively; and denote the predicted stability moment level and the actual stability moment level, is the static stability coefficient; is the characteristic energy factor; is the characteristic moment factor;

[0159] The height influence coefficient represents the influence of height, and the expression is:

[0160]

[0161] The stability weight coefficient evaluates the influence of stability, and the square of the predicted velocity reflects the influence of dynamic pressure. At the same time, the influence of environmental characteristics, parachute area, average aerodynamic chord length and characteristic length are considered. The expression is:

[0162] ;

[0163] S4.5. Construct a UAV aerodynamic deceleration control prediction model using a composite loss function consisting of speed loss, flight characteristics loss, and stability loss. The loss function is expressed as:

[0164]

[0165] in, 、 、 、 They are 、 、 、 The weight coefficient of .

[0166] S5. Based on the normalized input characteristic parameter data of the UAV aerodynamic deceleration control prediction model obtained in step S1 and the normalized output index data of the UAV aerodynamic deceleration control prediction model obtained in step S3, a training set, a validation set, and a test set are constructed, and parameters are set for the constructed UAV aerodynamic deceleration control prediction model. Then, the UAV aerodynamic deceleration control prediction model is trained using the obtained training set, verified using the validation set, and the verified UAV aerodynamic deceleration control prediction model is tested using the test set.

[0167] Furthermore, the data sources include commercial flight data obtained through the Civil Aviation Administration of China, test data obtained through large drone manufacturers, design parameters obtained through drone design manuals, experimental data obtained through drone laboratory tests, and drone monitoring data obtained through drone monitoring and management platforms.

[0168] Furthermore, in step S5, the input characteristic parameter data of the normalized UAV aerodynamic deceleration control prediction model and the output index data of the normalized UAV aerodynamic deceleration control prediction model obtained in step S3 are divided into a training set, a validation set, and a test set in a ratio of 8:1:1.

[0169] In step S5, the model training adopts the back propagation algorithm. During the training process, the model calculates the predicted value through forward propagation, substitutes the predicted value and the true value into the loss function to calculate the loss, and then updates the network parameters through back propagation.

[0170] In step S5, to prevent overfitting, an early stopping strategy is adopted, and training is stopped when the validation set loss does not decrease for 10 consecutive epochs. After the model training is completed, the relationship between input and output is established, and the output result is restored to the actual physical quantity through denormalization.

[0171] Furthermore, the model parameter settings need to be manually set based on expert experience. Hyperparameters include: network structure parameters, the number of neurons in each layer; hyperparameters that need to be manually set during the model training process include: learning rate, batch size, training rounds, and weight coefficients in the loss function; parameters that need to be determined by the model during training include: weight coefficients and bias terms of each layer.

[0172] Furthermore, by inputting X1~X 20 The corresponding parameters can be predicted to get Y1~Y 10 ; The output result is then restored to the actual physical quantity through denormalization.

[0173] Table 2 shows some input data of the prediction model of UAV aerodynamic deceleration control; Figure 2 This is a loss curve diagram of the training process of the UAV aerodynamic deceleration control prediction model of the present invention; Figure 3 This is a comparison chart of the predicted value and the actual value of the stability index of the UAV aerodynamic deceleration control prediction model of the present invention.

[0174] Table 2

[0175]

[0176] Example 2:

[0177] An electronic device includes a memory and a processor, the memory storing a computer program, and the processor implementing the steps of the unmanned aerial vehicle aerodynamic deceleration control prediction method described in any one of Example 1 when executing the computer program.

[0178] The computer device of the present invention may include a processor and memory, such as a single-chip microcomputer including a central processing unit (CPU). Furthermore, the processor is configured to execute a computer program stored in the memory to implement the steps of the aforementioned method for predicting aerodynamic deceleration control of a drone. The processor may be a central processing unit (CPU), other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA), other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor may be a microprocessor, or any conventional processor. The memory may primarily include a program storage area and a data storage area. The program storage area may store an operating system and at least one application required for a function (such as audio playback or image playback); the data storage area may store data generated based on the use of the mobile phone (such as audio data, a phone book, etc.). In addition, the memory may include high-speed random access memory and non-volatile memory, such as a hard disk, internal memory, a plug-in hard disk, a Smart Media Card (SMC), a Secure Digital (SD) card, a flash card, at least one disk storage device, a flash memory device, or other volatile solid-state storage device. Example 3:

[0179] A computer-readable storage medium having a computer program stored thereon, wherein when the computer program is executed by a processor, the method for predicting aerodynamic deceleration control of a drone as described in Example 1 is implemented.

[0180] The computer-readable storage medium of the present invention can be any form of storage medium readable by a processor of a computer device, including but not limited to non-volatile memory, volatile memory, ferroelectric memory, etc. The computer-readable storage medium stores a computer program. When the processor of the computer device reads and executes the computer program stored in the memory, the steps of the aforementioned method for predicting aerodynamic deceleration control of a drone can be implemented. The computer program includes computer program code, which can be in source code form, object code form, executable file, or some intermediate form. The computer-readable medium can include any entity or device capable of carrying the computer program code, a recording medium, a USB flash drive, a mobile hard drive, a magnetic disk, an optical disk, computer memory, read-only memory (ROM), random access memory (RAM), an electrical carrier signal, a telecommunications signal, and software distribution media.

[0181] It should be noted that relational terms such as "first" and "second" are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the terms "comprises," "comprising," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus comprising a series of elements includes not only those elements, but also other elements not explicitly listed, or elements inherent to such process, method, article, or apparatus. In the absence of further limitations, an element defined by the phrase "comprising a ..." does not exclude the presence of additional identical elements in the process, method, article, or apparatus comprising the element.

[0182] Although the present application has been described above with reference to specific embodiments, various modifications may be made thereto and components may be substituted with equivalents without departing from the scope of the present application. In particular, as long as there are no structural conflicts, the various features of the embodiments disclosed herein may be combined with each other in any manner, and the omission of an exhaustive description of these combinations in this specification is solely for the sake of space and resource conservation. Therefore, the present application is not limited to the specific embodiments disclosed herein, but includes all technical solutions within the scope of the claims.

Claims

1. A method for predicting aerodynamic deceleration control of a UAV, characterized in that: The steps include: S1. Collect and normalize the input characteristic parameters of the UAV aerodynamic deceleration control prediction model, including flight state parameters, control input parameters, and environmental parameters; S2. Construct a hidden layer for a UAV aerodynamic deceleration control prediction model using a three-layer architecture consisting of a dynamics perception layer, an aerodynamic characteristics layer, and a control decision layer. S3. Output indicators of the hidden layer of the UAV aerodynamic deceleration control prediction model obtained in step S2, including flight state prediction indicators and control effect evaluation indicators; flight state prediction indicators include predicted airspeed Y1, predicted angle of attack Y2, predicted altitude Y3, predicted Mach number Y4, and predicted dynamic pressure Y5; control effect evaluation indicators include deceleration rate Y6, aerodynamic drag coefficient Y7, lift-to-drag ratio Y8, energy loss rate Y9, and stability index Y 10 ;Establish the mapping relationship between the model hidden layer and output indicators; S4. Construct a loss function for the UAV aerodynamic deceleration control prediction model that is a composite of the base loss function, the speed loss term, the flight characteristics loss term, and the stability loss term. S5. Based on the normalized input characteristic parameter data of the UAV aerodynamic deceleration control prediction model obtained in step S1 and the normalized output index data of the UAV aerodynamic deceleration control prediction model obtained in step S3, a training set, a validation set, and a test set are constructed, and parameters are set for the constructed UAV aerodynamic deceleration control prediction model. Then, the UAV aerodynamic deceleration control prediction model is trained using the obtained training set, verified using the validation set, and tested using the verified UAV aerodynamic deceleration control prediction model.

2. The method for predicting aerodynamic deceleration control of a UAV according to claim 1, characterized in that: The input characteristic parameters of the UAV aerodynamic deceleration control prediction model in step S1 are as follows: Flight status parameters include airspeed X1, angle of attack X2, sideslip angle X3, altitude X4, Mach number X5, dynamic pressure X6, roll angle X7 and pitch angle X8; The control input parameters include aileron deflection angle X9, rudder deflection angle X 10 , elevator deflection angle X 11 , throttle opening X 12 , air brake plate deflection angle X 13 and the deployment area of the parachute X 14 ; Environmental parameters include air density X 15 , temperature X 16 , wind speed X 17 、wind direction X 18 , atmospheric pressure X 19 and humidity X 20 .

3. The method for predicting aerodynamic deceleration control of a UAV according to claim 2, characterized in that: The specific implementation method of step S2 includes the following steps: S2.

1. Based on the consideration of airspeed and altitude among flight state parameters, and the impact of environmental parameters on dynamic characteristics, a dynamic perception layer is constructed. The expression is: ； in, is the output of the dynamic perception layer, is the activation function, is the weight coefficient of the input characteristic parameters of the aerodynamic deceleration control prediction model of the i-th UAV, X i is the input characteristic parameter of the aerodynamic deceleration control prediction model of the i-th UAV, is the dynamic characteristic coefficient, is a high impact factor, is the characteristic time scale factor, which characterizes the characteristic response speed of the UAV during deceleration; is the characteristic height scale factor, which represents the sensitivity of the UAV to altitude changes; is the bias term of the dynamic perception layer, i=1~20; Dynamic characteristic coefficient The expression reflecting the dynamic characteristics of the UAV during deceleration is: ； in, For the quality of the drone, is the drag coefficient; High impact factor The relationship between altitude and atmospheric pressure is expressed as: ； in, is the reference height, is standard atmospheric pressure; S2.

2. Based on the dynamic characteristics, the effects of angle of attack and dynamic pressure on aerodynamic forces are considered, and the Mach number effect is considered to construct the aerodynamic characteristic layer. The expression is: ； in, Output for the aerodynamic characteristics layer; is the bias term of the aerodynamic characteristic layer, is the lift characteristic coefficient; is the compressibility effect factor, is the characteristic angle of attack, which characterizes the response characteristics of the system to attitude changes during deceleration; is the characteristic dynamic pressure, which represents the adaptability of the UAV to aerodynamic changes; lift characteristic coefficient The aerodynamic characteristics are reflected by the ratio of dynamic pressure, parachute deployment area and gravity, and the expression is: ； Where g is the acceleration due to gravity; Compressibility effect factor Considering the influence of air density and Mach number on the compressibility effect, the expression is: ； S2.

3. Construct a control decision layer. By establishing a direct information channel from the dynamics perception layer to the control decision layer, a rapid mapping of the initial dynamic state to the control decision is achieved. The expression is: ； in, To control the output of the decision-making layer; is the residual connection coefficient, determined by expert experience; is the control efficiency coefficient; is the stability influencing factor, To control the bias term of the decision layer; is the characteristic force, which represents the impact of the change of control force on the UAV; is the characteristic rudder deflection angle, which represents the effect of rudder surface deflection on the UAV; Control efficiency coefficient The control efficiency is reflected by considering the relationship between dynamic pressure and characteristic length. The product of dynamic pressure and the deployment area of the parachute represents the aerodynamic control force. The expression is: ； in, is the characteristic length, is the mean aerodynamic chord length; Stability influencing factors The stability evaluation index is constructed by Mach number, and the expression is: 。 4. The method for predicting aerodynamic deceleration control of a UAV according to claim 3, characterized in that: The specific implementation method of step S3 is to perform a weighted combination of the outputs of the three hidden layers constructed in step S2, and introduce a nonlinear activation function and a characteristic scale factor of the corresponding indicator to achieve accurate prediction of the flight status prediction indicator and the control effect evaluation indicator. The nonlinear activation function introduced by the flight status prediction indicator is the hyperbolic tangent function, and the nonlinear activation function introduced by the control effect evaluation indicator is the sigmoid function.

5. The method for predicting aerodynamic deceleration control of a UAV according to claim 4, characterized in that: The specific implementation method of step S4 includes the following steps: S4.

1. Design a basic loss function and use the mean square error to calculate all output indicators Y1~Y 10 The mean square error between the predicted value and the measured value is the basic loss ; S4.

2. Design a speed loss function. This function quantifies the deceleration effect by evaluating the difference between the predicted and actual airspeeds, as well as the corresponding state transition efficiency. This function also takes into account the effects of air density and the deployment area of the parachute. L 1; S4.

3. Design a flight performance loss function. By considering the impact of the Mach number effect on aerodynamic characteristics, evaluate the accuracy of lift, drag, and torque predictions, and introduce characteristic length and aerodynamic parameters to construct a comprehensive evaluation index for flight quality. The flight performance loss function L 2; S4.

4. Design a stability loss function. By evaluating three aspects, namely height change, static stability and dynamic stability, a complete stability evaluation system is constructed. The stability loss function L 3; S4.

5. Construct a UAV aerodynamic deceleration control prediction model using a composite loss function consisting of speed loss, flight characteristics loss, and stability loss. The loss function is expressed as: ； in, 、 、 、 They are 、 、 、 The weight coefficient of .

6. The method for predicting aerodynamic deceleration control of a UAV according to claim 5, characterized in that: In step S5, the input characteristic parameter data of the normalized UAV aerodynamic deceleration control prediction model and the output index data of the normalized UAV aerodynamic deceleration control prediction model obtained in step S3 are divided into a training set, a validation set, and a test set in a ratio of 8:1:

1.

7. The method for predicting aerodynamic deceleration control of a UAV according to claim 6, characterized in that: In step S5, the model training adopts the back propagation algorithm. During the training process, the model calculates the predicted value through forward propagation, substitutes the predicted value and the true value into the loss function to calculate the loss, and then updates the network parameters through back propagation.

8. The method for predicting aerodynamic deceleration control of a UAV according to claim 7, characterized in that: In step S5, to prevent overfitting, an early stopping strategy is adopted, and training is stopped when the validation set loss does not decrease for 10 consecutive epochs. After the model training is completed, the relationship between input and output is established, and the output result is restored to the actual physical quantity through denormalization.

9. An electronic device, characterized in that: The invention comprises a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, the steps of the method for predicting the aerodynamic deceleration control of a drone according to any one of claims 1 to 8 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the method for predicting the aerodynamic deceleration control of a drone according to any one of claims 1 to 8 is implemented.

Images