Composite wind field environment unmanned aerial vehicle anti-disturbance control method without airflow angle measurement

Abstract

The application discloses a composite wind field environment unmanned aerial vehicle anti-disturbance control method without airflow angle measurement, and belongs to the field of unmanned aerial vehicle navigation, guidance and control. Firstly, an affine nonlinear processing is performed on a six-degree-of-freedom dynamic model of an unmanned aerial vehicle, and an affine nonlinear motion model of the unmanned aerial vehicle is established. Then, based on deep learning, an angle conversion method between airflow angles and Euler angles of the unmanned aerial vehicle under wind disturbance conditions is designed, and airflow angle information of the unmanned aerial vehicle can be accurately estimated without airflow angle measurement sensors. Finally, based on the affine nonlinear motion model of the unmanned aerial vehicle and the angle conversion method under wind disturbance conditions, an overall anti-disturbance control framework of the unmanned aerial vehicle from an attitude loop to a position loop is constructed, and an anti-disturbance control law of the unmanned aerial vehicle is designed, so that the anti-disturbance control capability of the unmanned aerial vehicle is improved under the conditions of no airflow angle measurement and composite wind field disturbance.

Description

Technical Field

[0001] This invention belongs to the field of unmanned aerial vehicle (UAV) navigation, guidance and control, and specifically relates to a method for anti-disturbance control of UAVs in composite wind field environments without airflow angle measurement. Background Technology

[0002] In actual flight, drones are subject to disturbances from complex wind fields (constant wind, atmospheric turbulence, gusts, etc.) and aerodynamic parameter perturbations, exhibiting characteristics such as rapid time-varying, strong nonlinearity, and strong coupling. Furthermore, small drones are difficult to install complex atmospheric data systems, making direct measurement of airflow angles challenging. In fact, airflow angles (especially angles of attack) are crucial state information for drones; drones must constantly monitor and understand airflow angle states to ensure flight safety. Therefore, airflow angles are essential information for safe drone flight. In addition, when performing missions, drones need to be controlled in conjunction with constrained Euler angles.

[0003] Current active position control methods for UAVs are mostly linear control-based, such as linear quadratic optimal control (LQR), which cannot effectively address the challenges of multiple disturbances, strong nonlinearity, and strong coupling in UAV position control. Existing disturbance rejection control methods, such as active disturbance rejection control (ADRC), all require airflow angle measurement information from the UAV. When airflow angle measurement is unavailable, the entire control framework fails, making it difficult to stabilize the UAV's wobbling motion under windy conditions. Although some literature uses Kalman filtering to estimate the UAV's airflow angle state, this method relies on accurate aerodynamic model data, and the estimation accuracy decreases significantly when aerodynamic parameters are perturbed. Summary of the Invention

[0004] This invention addresses the control challenges of UAVs, such as strong coupling, strong nonlinearity, inability to measure airflow angles, and the need for constrained control of Euler angles, which are difficult to solve with traditional control methods. To further improve the position control performance of UAVs in complex wind field environments with disturbances and parameter perturbations, this invention proposes a disturbance-resistant control method for UAVs in complex wind field environments without airflow angle measurement.

[0005] The aforementioned method for anti-disturbance control of unmanned aerial vehicles (UAVs) in composite wind field environments without airflow angle measurement specifically includes the following steps:

[0006] Step 1: Perform affine nonlinear processing on the six-degree-of-freedom dynamic model of the UAV to establish an affine nonlinear motion model of the UAV.

[0007] The affine nonlinear motion model of the UAV is as follows:

[0008]

[0009] Among them, Xb =[y b ,z b ] T Let y be the state variable of the centroid loop. b The lateral position of the drone, z b F represents the vertical position of the drone. b For the lumped disturbance of the centroid loop, B b The control matrix for the centroid loop; Here, χ represents the state variable of the track angle loop, γ represents the track deflection angle, and γ represents the track inclination angle. This refers to the lumped disturbance of the track angle loop. Here is the control matrix for the track angle loop, υ = [υ1, υ2]. T X is an intermediate variable; a =[α,β,μ] T α is the angle of attack, β is the sideslip angle, and μ is the velocity roll angle; w ,β w X represents the angular component of the airflow caused by wind disturbance. Ω =[ψ,θ,φ] T and X ω =[p,q,r] T Let F be the state variables of the attitude angle loop and the angular velocity loop, respectively; ψ, θ, φ be the yaw angle, pitch angle, and roll angle of the UAV, respectively; and p, q, r be the roll angular velocity, pitch angular velocity, and yaw angular velocity of the UAV, respectively. ω For the lumped disturbance of the angular velocity loop, B Ω and B ω These are the control matrices for the attitude angle loop and the angular velocity loop, respectively, where δ is the aerodynamic control input; V k For the ground speed of the drone, and These are the lumped disturbance and control input coefficient of the speed loop, δ, respectively. T f is the throttle opening. a2Ω (·) represents the airflow angle X a To Euler angle X Ω Angle conversion function.

[0010] Step 2: Based on deep learning, design an angle conversion method between airflow angle and Euler angle under wind disturbance conditions to establish an affine connection between the position loop and the attitude loop.

[0011] The specific steps are as follows:

[0012] Step 201: Based on the conversion relationship between the inertial frame and the machine frame, obtain the Euler angles and airflow angular components α of the UAV. k and β k The conversion relationship between them;

[0013] The airflow angles α and β of a drone consist of two parts: one part is α caused by wind disturbance. w and β w The other part is α caused by the trajectory speed. k and β k ,Right now:

[0014]

[0015] α k and β k The Euler angles φ, θ, ψ of the UAV can be obtained by converting them from the inertial frame and the machine frame based on the conversion relationship between them:

[0016] R x (φ)R y (θ)R z (ψ)=R y (α k )R z (-β k )R x (μ)R y (γ)R z (χ) (3)

[0017] Among them, R x (φ) represents the transformation matrix with a rotation angle of φ around the X-axis, and the definitions of other symbols are similar.

[0018] Step 202: Design a deep learning-based airflow angle estimation network;

[0019] Based on deep learning methods, a proposal including N L A flow angle estimation network consisting of one LSTM layer and one fully connected layer.

[0020] The input S of the airflow angle estimation network d The design is as follows:

[0021]

[0022] Network input S d The total disturbance includes estimates from the track angle loop and angular velocity loop ESOs. and The states of the UAV's trajectory angle loop, Euler angle loop, and angular velocity loop X Ω and X ω The airspeed V and ground speed V of the drone k (All are scalars.)

[0023] The forward propagation process of the airflow angle estimation network is defined by the following function:

[0024]

[0025] Among them, f AEN (·) is the forward propagation function of the airflow angle estimation network.

[0026] The training method for the airflow angle estimation network is as follows: input the randomly given atmospheric turbulence intensity, the parameter perturbation of the model, and the reference trajectory of the controller into the simulation model to obtain training samples, and then train the airflow angle estimation network using the training samples.

[0027] The simulation model includes a composite wind field model, a UAV dynamics model, and a UAV anti-disturbance position controller. The UAV position controller is designed with active disturbance rejection control and assumes that the airflow angle can be accurately measured.

[0028] Step 203: Based on the conversion relationship between the inertial frame and the machine frame and the airflow angle estimation network, the conversion relationship between the airflow angle and the Euler angle under the combined wind field is obtained:

[0029] First, the flight path angle α is obtained through coordinate transformation based on the current Euler angles and flight path angles of the UAV. k ,β k Further combining the airflow angle data obtained from the airflow angle estimation network yields... As shown in equation (6).

[0030]

[0031] Then, based on the airflow angle command X obtained from the track angle controller a * =[α * β * μ * ] T and Calculate the desired flight path airflow angle command Further combined with track angle commands Euler angle command X is obtained from the coordinate system transformation relationship (generated by the position loop controller) and the coordinate system transformation relationship. Ω * ,Right now The calculation process is shown in equation (7).

[0032]

[0033] Therefore, based on the airflow angle obtained by the airflow angle estimation network, the commands generated by the track loop can be converted into command signals of the attitude loop through equations (6) and (7), thereby establishing an affine connection between the position loop and the attitude loop.

[0034] Step 3: Based on the affine nonlinear motion model of the UAV and the angle conversion method under wind disturbance conditions, construct an anti-disturbance control framework for the UAV under the condition of no airflow angle measurement, design the anti-disturbance control law of the UAV, and realize the anti-disturbance control of the UAV under the conditions of no airflow angle measurement and complex wind field disturbance.

[0035] According to the disturbance rejection control framework, the disturbance rejection control laws of the UAV include, in sequence, the position loop control law, the track angle loop control law, the Euler angle loop control law, the angular velocity loop control law, and the ground speed loop control law. The specific steps are as follows:

[0036] Step 301: Define the position tracking error as e b =X b * -X b Based on the affine nonlinear motion model of the UAV, a position loop control law is designed.

[0037]

[0038] Among them, X b * For the desired drone location, K b The parameters for the position loop controller to be designed are as follows: The position loop lumped disturbance is estimated by the Extended State Observer (ESO). X is estimated by the tracking differentiator (TD). b * First-order differential value.

[0039] Step 302: Define the track angle tracking error as... Design the track angle loop control law;

[0040]

[0041] in, The parameters for the track angle loop controller to be designed are as follows: The track angle loop lumped disturbance estimated by ESO. For the estimated by TD First-order differential value.

[0042] Furthermore, based on the affine nonlinear motion model of the UAV, the trajectory angle loop control law υ is applied. * Convert to airflow angle command:

[0043]

[0044] Step 303: Convert the airflow angle command generated by the track angle loop into the Euler angle command of the attitude loop, and define the Euler angle tracking error e. Ω Design the Euler angle loop control law;

[0045]

[0046] Among them, X Ω * The Euler angles are the commands obtained from the conversion. X obtained from TD Ω * First-order differential value, K Ω e represents the gain parameters of the attitude angle loop controller to be designed. υ =υ * -υ.

[0047] Step 304: Define the angular velocity loop tracking error as e. ω =X ω * -X ω Design the angular velocity loop control law;

[0048]

[0049] Among them, K ω The parameters for the angular velocity loop controller to be designed are as follows: For the lumped disturbance of the angular velocity loop estimated by ESO, X estimated by TD ω * First-order differential value.

[0050] Step 305: Define the speed tracking error as... Design the ground speed loop control law;

[0051]

[0052] Among them, V k * For speed commands, The parameters for the speed loop controller to be designed are as follows: The velocity loop lumped disturbance estimated by ESO, V estimated by TD k * First-order differential value.

[0053] The advantages of this invention are:

[0054] (1) A method for anti-disturbance control of UAV in composite wind field environment without airflow angle measurement, which is based on deep learning to accurately estimate the airflow angle of UAV in real time, and can accurately estimate the airflow angle information of UAV without airflow angle measurement sensor.

[0055] (2) A disturbance-resistant control method for UAVs in a composite wind field environment without airflow angle measurement, which connects the position loop and attitude loop based on the angle conversion method, establishes an overall disturbance-resistant control framework for UAVs from attitude loop to position loop, and improves the disturbance-resistant capability of UAV position control.

[0056] (3) A method for anti-disturbance control of UAV in composite wind field environment without airflow angle measurement, which can realize direct control of Euler angle based on angle conversion method. Attached Figure Description

[0057] Figure 1 This is a schematic diagram of the airflow angle composition of the UAV of the present invention;

[0058] Figure 2 This is a diagram illustrating the airflow angle estimation network structure and training method of the present invention;

[0059] Figure 3 This is a block diagram of the anti-disturbance control system for the UAV of the present invention;

[0060] Figure 4 This is a schematic diagram of wind disturbance information of a composite wind field in an embodiment of the present invention;

[0061] Figure 5 This is a graph showing the estimation performance of the airflow angle estimation network under parameter perturbation in an embodiment of the present invention.

[0062] Figure 6 This is a graph showing the estimation performance of the airflow angle estimation network under +20% parameter perturbation in an embodiment of the present invention.

[0063] Figure 7 This is a graph showing the estimation performance of the airflow angle estimation network under -20% parameter perturbation in an embodiment of the present invention.

[0064] Figure 8 This is a diagram illustrating the anti-disturbance trajectory tracking performance of an unmanned aerial vehicle (UAV) under +20% parameter perturbation conditions according to an embodiment of the present invention.

[0065] Figure 9 This is a diagram illustrating the anti-disturbance trajectory tracking performance of an unmanned aerial vehicle (UAV) under -20% parameter perturbation conditions according to an embodiment of the present invention. Detailed Implementation

[0066] To facilitate understanding and implementation of the present invention, the present invention will be further described in detail below with reference to the accompanying drawings and examples.

[0067] Due to the unavailability of airflow angle measurements and the need for attitude control in UAVs, the inner loop must use Euler angles as the control variable, leading to a break in the affine connection between the position and attitude loops, making it difficult to establish an overall disturbance-resistant control framework. This invention aims to achieve real-time and accurate estimation of the airflow angle state of a UAV in a complex wind field environment without airflow angle measurement. It establishes an affine connection between the position and attitude loops through an angle conversion method under wind disturbance conditions, thereby constructing an overall disturbance-resistant control framework. This invention proposes a disturbance-resistant control method for UAVs in complex wind field environments without airflow angle measurement. First, real-time and accurate estimation of the airflow angle is performed based on deep learning, and an accurate conversion method from airflow angle to Euler angle is established by combining the conversion relationship between the inertial frame and the machine frame. Based on this angle conversion method, the position and attitude loops are connected, establishing an overall disturbance-resistant control framework for the UAV from attitude to position. This invention is of great significance for realizing disturbance-resistant control of small UAVs without airflow angle measurement equipment.

[0068] A disturbance-resistant control method for unmanned aerial vehicles (UAVs) in complex wind field environments without airflow angle measurement includes the following steps:

[0069] Step 1: Establish an affine nonlinear motion model for the UAV;

[0070] To design an anti-disturbance position control law for a UAV, the six-degree-of-freedom dynamic model of the UAV needs to be subjected to affine nonlinear processing first. UAVs typically employ roll-to-side-pinch cancellation, therefore the expected sideslip angle β... * =0. The ground speed of the UAV is directly controlled by changing the thrust of the engine through a separate control channel, thereby controlling the forward position x. b And the lateral position y b and vertical position z b Indirect control is achieved by changing the deflection angle of the UAV's aerodynamic control surfaces. However, considering the complexity of airflow angle measurement sensor systems and the difficulty of installing them on small UAVs, the UAV's airflow angle information cannot be directly measured. Furthermore, Euler angles of the UAV are typically constrained for control, therefore, the UAV's attitude motion needs to be described using Euler angles as the state.

[0071] Based on the analysis, the affine nonlinear motion model of the UAV can be written as follows:

[0072]

[0073] Among them, X b =[y b ,z b ] T F is the state variable of the centroid loop. b For the lumped disturbance of the centroid loop, B b The control matrix for the centroid loop; Here, χ represents the state variable of the track angle loop, γ represents the track deflection angle, and γ represents the track inclination angle. This refers to the lumped disturbance of the track angle loop. Here is the control matrix for the track angle loop, υ = [υ1, υ2]. T X is an intermediate variable; a =[α,β,μ] T α is the angle of attack, β is the sideslip angle, and μ is the velocity roll angle; w ,β w X represents the angular component of the airflow caused by wind disturbance. Ω =[ψ,θ,φ] T and X ω =[p,q,r] T Let F be the state variables of the attitude angle loop and the angular velocity loop, respectively; ψ, θ, φ be the yaw angle, pitch angle, and roll angle of the UAV, respectively; and p, q, r be the roll angular velocity, pitch angular velocity, and yaw angular velocity of the UAV, respectively. ω For the lumped disturbance of the angular velocity loop, B Ω and B ω These are the control matrices for the attitude angle loop and the angular velocity loop, respectively, where δ is the aerodynamic control input; V k For the ground speed of the drone, and These are the lumped disturbance and control input coefficient of the speed loop, δ, respectively. T f is the throttle opening. a2Ω (·) represents the airflow angle X a To Euler angle X Ω The angle transformation function will be designed in subsequent steps.

[0074] Step 2: Based on deep learning, design an angle conversion method between airflow angle and Euler angle under wind disturbance conditions to establish an affine connection between the position loop and the attitude loop, and construct an overall anti-disturbance control framework under the condition of no airflow angle measurement.

[0075] The specific steps are as follows:

[0076] Step 201: Based on the conversion relationship between the inertial frame and the machine frame, obtain the Euler angles and airflow angular components α of the UAV. k and β k The conversion relationship between them;

[0077] The airflow angles α and β of a drone consist of two parts: one part is α caused by wind disturbance. w and β w The other part is α caused by the trajectory speed. k and β k ;α k and β kDetermines the trajectory speed V k Relative to the orientation of the body, they are similar to α and β, but have no aerodynamic significance, such as Figure 1 As shown, in the left-hand angle of attack diagram, V b V kb and V wb Airspeed V and trajectory speed V are respectively k Wind speed V w The projection onto the plane of symmetry of the organism. Therefore:

[0078]

[0079] α k and β k The Euler angles φ, θ, ψ of the UAV can be obtained by converting them from the inertial frame and the machine frame based on the conversion relationship between them:

[0080] R x (φ)R y (θ)R z (ψ)=R y (α k )R z (-β k )R x (μ)R y (γ)R z (χ) (3)

[0081] Among them, R x (φ) represents the transformation matrix with a rotation angle of φ around the X-axis, and the definitions of other symbols are similar.

[0082] Step 202: Design a deep learning-based method for estimating airflow angles α and β;

[0083] Based on deep learning methods, an airflow angle estimation network is proposed, employing a long short-term memory (LSTM) network as its structure. The structure and training method of the airflow angle estimation network are as follows: Figure 2 As shown.

[0084] Network input S d The design is as follows:

[0085]

[0086] Network input S d The total disturbance includes estimates from the track angle loop and angular velocity loop ESOs. and Because the impact of wind disturbance on the UAV's airflow angle is directly reflected in the disturbances of these two loops; including the states of the UAV's trajectory angle loop, Euler angle loop, and angular velocity loop. X dΩ and X dω It also includes the airspeed V and ground speed V of the drone. k (Both are scalars), because the difference between airspeed and ground speed reflects information about wind disturbance.

[0087] The airflow angle estimation network structure contains N L The LSTM network consists of one LSTM layer and one fully connected layer. It comprises three "gates": an input gate (i), an output gate (o), and a forget gate (f). The input gate controls the extent to which the current input flows into the cell state m, the output gate determines the extent to which the cell state flows out as the layer's output, and the forget gate limits the extent to which the cell state from the previous time step becomes the cell state at the current time step. Through this three-gate structure, LSTM solves the long-term dependency problem inherent in traditional recurrent neural networks (RNNs). The forward propagation process of the airflow angle estimation network is as follows:

[0088]

[0089]

[0090] Where W represents the weight matrix, b represents the bias vector, and σ represents the sigmoid function (gating function). Let m represent the updated cell state, h represent the output vector of the LSTM layer, l represent the current input, t represent time, and ⊙ represent the Hadamard product (element-wise multiplication). K represents the output vector of the fully connected layer. s and C p It is a non-trainable constant used to... Shift and scale to a reasonable range to limit the output. and Right now

[0091]

[0092] Then, the forward propagation process of the airflow angle estimation network can be defined as the following function:

[0093]

[0094] Among them, f AEN (·) is the forward propagation function of the network.

[0095] Training methods for airflow angle estimation networks are as follows: Figure 2As shown. Training data is generated through a simulation model, which includes a UAV dynamics model, a composite wind field model, and a UAV anti-disturbance position controller (the UAV position controller is designed with active disturbance rejection control and assumes that the airflow angle can be accurately measured). By randomly assigning atmospheric turbulence intensity, model parameter perturbations, and the controller's reference trajectory, a large number of training samples are obtained to train the airflow angle estimation network.

[0096] Step 203: Based on the conversion relationship between the inertial frame and the machine frame and the airflow angle estimation network, the conversion relationship between the airflow angle and the Euler angle under the combined wind field is obtained:

[0097]

[0098]

[0099] First, the flight path angle α can be obtained from the current Euler angles and flight path angles of the UAV through coordinate transformation. k ,β k Furthermore, by combining the airflow angle data obtained from the airflow angle estimation network, we can obtain... As shown in equation (8).

[0100] Then, based on the airflow angle command X obtained from the trajectory loop controller a * =[α * β * μ * ] T and The desired flight path airflow angle command can be calculated. Further combined with track angle commands The Euler angle command X can be obtained from the coordinate system transformation relationship (generated by the position loop controller) and the coordinate system transformation relationship. Ω * ,Right now The calculation process is shown in equation (9).

[0101] Based on the estimated airflow angle obtained from the network, the commands generated by the track loop can be converted into command signals of the attitude loop through equations (8) and (9), thereby establishing a "bridge" connecting the inner and outer loops.

[0102] Based on the angle transformation method described above, an affine connection can be established between the position loop and the attitude loop.

[0103] Step 3: Based on the affine nonlinear motion model of the UAV and the angle conversion method under wind disturbance conditions designed above, construct an anti-disturbance control framework for the UAV under the condition of no airflow angle measurement, design anti-disturbance control laws for position loop, track angle loop, Euler angle loop, angular velocity loop and ground speed loop, and realize active anti-disturbance control of UAV position under adverse conditions such as the inability to measure airflow angle and complex wind field disturbance.

[0104] Disturbance-resistant control framework for UAVs in the absence of airflow angle measurement, such as Figure 3 As shown, the specific steps of the disturbance rejection control law are as follows:

[0105] Step 301: Design the position loop control law;

[0106] Define the position tracking error as e b =X b * -X b , where X b * To determine the desired UAV position, and based on the established affine nonlinear motion model of the UAV, the following position loop control law is designed:

[0107]

[0108] Among them, K b The parameters for the position loop controller to be designed are as follows: The position loop lumped disturbance is estimated by the Extended State Observer (ESO). X estimated by TD b * First-order differential value.

[0109] Step 302: Design the track angle loop control law;

[0110] Define the track angle tracking error as The control law for designing the trajectory angle loop is:

[0111]

[0112] in, The parameters for the track angle loop controller to be designed are as follows: The track angle loop lumped disturbance estimated by ESO. For the estimated by TD First-order differential value.

[0113] Based on the established affine nonlinear motion model of the UAV, the υ can be further... * Convert to airflow angle command

[0114]

[0115] Step 303: Design the Euler angle loop control law;

[0116] In actual flight, it is usually necessary to constrain the Euler angles of the UAV, thus requiring direct control of the Euler angle loop. This invention establishes a "bridge" connecting the inner and outer loops, that is, according to step two, the airflow angle command generated by the trajectory angle loop can be converted into the command of the attitude (Euler angle) loop. Then, the Euler angle tracking error e is defined. Ω And design the control law X of the Euler angle circuit. ω * as follows:

[0117]

[0118] Among them, X Ω * The Euler angles are the commands obtained from the conversion. X obtained from TD Ω * First-order differential value, K Ω e represents the gain parameters of the attitude angle loop controller to be designed. υ =υ * -υ.

[0119] Step 304, Angular velocity loop control law;

[0120] Define the angular velocity loop tracking error as e ω =X ω * -X ω The control law for the angular velocity loop is designed as follows:

[0121]

[0122] Among them, K ω The parameters for the angular velocity loop controller to be designed are as follows: For the lumped disturbance of the angular velocity loop estimated by ESO, X estimated by TD ω * First-order differential value.

[0123] Step 305, Ground speed loop control law;

[0124] This invention treats ground speed as a separate circuit and controls it through throttle opening, defining the speed tracking error as e. Vk =V k * -V k V k * For speed commands, the following speed control law is designed:

[0125]

[0126] in, The parameters for the speed loop controller to be designed are as follows: The velocity loop lumped disturbance estimated by ESO, V estimated by TD k * First-order differential value.

[0127] Example

[0128] To verify the effectiveness of this invention, a simulation was conducted using a certain type of UAV as an example. The controller parameters are shown in Table 1:

[0129] Table 1

[0130]

[0131] According to the specific implementation steps of this invention, firstly, composite wind field disturbance information including constant wind, atmospheric turbulence, and gusts is given, such as... Figure 4 As shown.

[0132] Then, the airflow angle estimation results obtained by the airflow angle estimation network are presented under the conditions of no parameter perturbation and parameter perturbation ±20% for the composite wind field, respectively. Figures 5-7 As shown, the airflow angle estimation network proposed in this invention can approximate the true value very well and has high estimation accuracy.

[0133] Furthermore, Table 2 presents the mean absolute error (MAE) results of the airflow angle estimation network. It can be seen that the airflow angle estimation network proposed in this invention still has high estimation accuracy under parameter perturbation, which fully verifies the performance of the airflow angle estimation network proposed in this invention.

[0134] Table 2

[0135] Parameter perturbation -20% Parameter perturbation 0% Parameter perturbation +20% Angle of attack 0.0401° 0.0338° 0.0329° Sideslip angle 0.0252° 0.0219° 0.0249°

[0136] Next, the positional anti-disturbance tracking performance of the UAV under complex wind field disturbances and parameter uncertainties was verified, such as... Figure 8 and Figure 9 As shown, even without airflow angle measurement, the method proposed in this invention can still resist adverse factors such as complex wind fields and parameter perturbations, thereby accurately tracking the command trajectory.

[0137] The simulation verification of the above embodiments proves the effectiveness of the anti-disturbance control method for UAVs in composite wind field environments without airflow angle measurement.

[0138] The contents not described in detail in this specification are existing technologies known to those skilled in the art.

Claims

1. A method for anti-disturbance control of unmanned aerial vehicles (UAVs) in composite wind field environments without airflow angle measurement, characterized in that, Specifically as follows: First, the six-degree-of-freedom dynamic model of the UAV is subjected to affine nonlinear processing to establish the affine nonlinear motion model of the UAV. Then, based on deep learning, an angle transformation method between the airflow angle and Euler angle of the UAV under wind disturbance conditions is designed to establish an affine connection between the position loop and the attitude loop; as detailed below: The flight path angle α is obtained through coordinate transformation based on the current Euler angles and flight path angles of the UAV. k ,β k Furthermore, by combining the airflow angle data obtained from the airflow angle estimation network, the following calculations were performed: As shown in equation (1): In the formula, χ is the track deflection angle, γ is the track tilt angle, and ψ, θ, φ are the yaw angle, pitch angle, and roll angle of the UAV, respectively. Based on the airflow angle command X obtained from the track angle controller a * =[α * β * μ * ] T and Calculate the desired flight path airflow angle command Further combined with track angle commands The Euler angle command X is obtained from the coordinate transformation relationship. Ω * ,Right now The calculation process is shown in equation (2): Therefore, based on the airflow angle obtained by the airflow angle estimation network, the command generated by the track loop can be converted into the command signal of the attitude loop through equations (1) and (2), thereby establishing an affine connection between the position loop and the attitude loop. Finally, based on the affine nonlinear motion model of the UAV and the angle conversion method under wind disturbance conditions, an anti-disturbance control framework for the UAV under the condition of no airflow angle measurement is constructed, and the anti-disturbance control law of the UAV is designed to realize the anti-disturbance control of the UAV under the conditions of no airflow angle measurement and complex wind field disturbance. The disturbance rejection control laws for UAVs include, in sequence, the position loop control law, the trajectory angle loop control law, the Euler angle loop control law, the angular velocity loop control law, and the ground speed loop control law. The specific steps are as follows: Step 301: Define the position tracking error as e b =X b * -X b Based on the affine nonlinear motion model of the UAV, a position loop control law is designed. Among them, X b * For the desired drone location, X b B is the state variable of the centroid loop. b K is the control matrix of the centroid loop. b The parameters for the position loop controller to be designed are as follows: The position loop lumped disturbance estimated by ESO. X estimated by TD b * First-order differential value; Step 302: Define the track angle tracking error as... Design the track angle loop control law; in, The control matrix for the track angle loop. The parameters for the track angle loop controller to be designed are as follows: The track angle loop lumped disturbance estimated by ESO. For the estimated by TD First-order differential value; Furthermore, based on the affine nonlinear motion model of the UAV, the trajectory angle loop control law υ is applied. * Convert to airflow angle command: in, Step 303: Convert the airflow angle command generated by the track angle loop into the Euler angle command of the attitude loop, and define the Euler angle tracking error e. Ω Design the Euler angle loop control law; Among them, X Ω * The Euler angles are the commands obtained from the conversion. B is the angle transformation function from airflow angle to Euler angle. Ω The control matrix for the attitude angle loop. X obtained from TD Ω * First-order differential value, K Ω e represents the gain parameters of the attitude angle loop controller to be designed. υ =υ * -υ; Step 304: Define the angular velocity loop tracking error as e. ω =X ω * -X ω Design the angular velocity loop control law; Among them, B ω K is the control matrix for the angular velocity loop. ω The parameters for the angular velocity loop controller to be designed are as follows: For the lumped disturbance of the angular velocity loop estimated by ESO, X estimated by TD ω * First-order differential value; Step 305: Define the speed tracking error as... Design the ground speed loop control law; Among them, V k * For speed commands, The parameters for the speed loop controller to be designed are as follows: The velocity loop lumped disturbance estimated by ESO, V estimated by TD k * First-order differential value.

2. The method for anti-disturbance control of a UAV in a composite wind field environment without airflow angle measurement according to claim 1, characterized in that, The affine nonlinear motion model of the UAV is as follows: Among them, X b =[y b ,z b ] T Let y be the state variable of the centroid loop. b The lateral position of the drone, z b F represents the vertical position of the drone. b For the lumped disturbance of the centroid loop, B b The control matrix for the centroid loop; Here, χ represents the state variable of the track angle loop, γ represents the track deflection angle, and γ represents the track inclination angle. For the lumped disturbance of the track angle loop, B φ Here is the control matrix for the track angle loop, υ = [υ1, υ2]. T X is an intermediate variable; a =[α,β,μ] T α is the angle of attack, β is the sideslip angle, and μ is the velocity roll angle; w ,β w X represents the angular component of the airflow caused by wind disturbance. Ω =[ψ,θ,φ] T and X ω =[p,q,r] T Let F be the state variables of the attitude angle loop and the angular velocity loop, respectively; ψ, θ, φ be the yaw angle, pitch angle, and roll angle of the UAV, respectively; and p, q, r be the roll angular velocity, pitch angular velocity, and yaw angular velocity of the UAV, respectively. ω For the lumped disturbance of the angular velocity loop, B Ω and B ω These are the control matrices for the attitude angle loop and the angular velocity loop, respectively, where δ is the aerodynamic control input; V k For the ground speed of the drone, F Vk and B Vk These are the lumped disturbance and control input coefficient of the speed loop, δ, respectively. T f is the throttle opening. a2Ω (·) represents the airflow angle X a To Euler angle X Ω Angle conversion function.

3. The method for anti-disturbance control of a UAV in a composite wind field environment without airflow angle measurement according to claim 2, characterized in that, The airflow angles α and β consist of two parts: Part of it is caused by wind disturbance. w and β w The other part is α caused by the trajectory speed. k and β k ,Right now: α k and β k The Euler angles φ, θ, ψ of the UAV can be obtained by converting them from the inertial frame and the machine frame based on the conversion relationship between them: R x (φ)R y (i)R z (ψ)=R y (a k )R z (-b k )R x (μ)R y (c)R z (x) (10) Among them, R x (φ) represents the transformation matrix with a rotation angle of φ around the X-axis, and the definitions of other symbols are similar.

4. The method for anti-disturbance control of a UAV in a composite wind field environment without airflow angle measurement according to claim 1, characterized in that, The airflow angle estimation network includes N L One LSTM layer and one fully connected layer; The input S of the airflow angle estimation network d The design is as follows: Network input S d The total disturbance includes estimates from the track angle loop and angular velocity loop ESOs. and The states of the UAV's trajectory angle loop, Euler angle loop, and angular velocity loop X Ω and X ω The airspeed V and ground speed V of the drone k ; The forward propagation process of the airflow angle estimation network is defined by the following function: Among them, f AEN (·) is the forward propagation function of the airflow angle estimation network.

5. The method for anti-disturbance control of a UAV in a composite wind field environment without airflow angle measurement according to claim 4, characterized in that, The training method for the airflow angle estimation network is as follows: input the randomly given atmospheric turbulence intensity, the parameter perturbation of the model, and the reference trajectory of the controller into the simulation model to obtain training samples, and train the airflow angle estimation network using the training samples.

6. The method for anti-disturbance control of a UAV in a composite wind field environment without airflow angle measurement according to claim 5, characterized in that, The simulation model includes a composite wind field model, a UAV dynamics model, an airflow angle estimation network, angle conversion relationships, and a UAV anti-disturbance position controller.

Images